Professional Documents
Culture Documents
ncir
Deresi around ayrkazan located northeast of Isklar
in the middle elevations of the mountain range between
the northernmost remains of the aqueducts at about 800
masl and the high ridge of the Baba Dag range hun-
dreds of meters above. Alternatively, if the aqueducts
were much longer, they could have collected water from
the numerous springs around Yayla Aras northwest of
Isklar, or from the springs just west of Tekam that feed
into the Dem Deresi north of Isklar.
After tapping these springs, both aqueducts traveled
from the I
ncir Deresi,
looking south (B081).
F
i
g
u
r
e
8
.
K
a
v
a
k
l
D
e
r
e
b
r
i
d
g
e
o
f
I
s
k
l
a
r
a
q
u
e
d
u
c
t
(
A
0
0
8
)
(
1
:
2
0
0
)
.
THE AQUEDUCTS OF APHRODISIAS 249
and a projected height above the streambed of 11.5 m.
The original bridge may have been more than 42 m long
and 15 m high.
The facing blocks and the stones of the rubble core
vary greatly in size, and most are fashioned from the
same locally available, garnet-embedded schist into
which the upstream tunnel was dug. In contrast, almost
all the external voussoirs are marble. Courses of blocks
of standard size are interchanged with courses of
smaller, more irregular stones, with an extensive use of
chinking stones. The lime mortar used throughout the
structure is pale gray-tan, with densely packed stone in-
clusions no larger than 0.01 m.
Where the pier is closest to the stream, large blocks
form a foundation resting on a rock outcrop. The arch
springs from a point 3.6 m above the base, and the three
lowest courses of voussoirs are recessed 0.15 m from the
short (south) side of the pier facing the stream. The
other voussoirs are recessed an additional 0.15 m, so that
the resulting double-stepped profle of the arch provides
two ledges that could have supported a wooden frame-
work for construction of the arch. Above this upper re-
cess, the pier narrows in width by 0.10 m on either side
above a course of large facing stones. These lateral
ledges extend even to the voussoirs of the arch spanning
the stream, with the result that the lowest three courses
of the arch are wider than the rest. In contrast, the arch
on the north side of the pier has no ledges to support a
framework, but a putlog hole just above the springing of
the arch fulflls the same purpose. It is noteworthy that
both arches were equipped with support for a frame-
work at a point that is three voussoirs above the spring
point, and that two different methods were used.
Between the bridge over the Kavakl Dere and
Aphrodisias, the aqueduct crossed the Derince Dere on
a small bridge of which only a masonry wall remains
(E050) (Fig. 10). The wall is constructed of mortared
rubble faced with petit appareil masonry and is built
into the east (left) bank of the stream. The top of the
wall is fush with the ground level of the stream bank,
suggesting that the structure has been leveled off dur-
ing centuries of agricultural activity in the surrounding
felds. The wall is 0.80 m wide and extends 2 m below
the current ground level of the stream bank.
The masonry includes feldstones of quartz, schist,
and sandstone varying greatly in size and bonded with
a gray-tan lime mortar with stone inclusions no larger
than 0.01 m. A possible construction trench, 0.70 m
wide, is discernible next to the south (downstream) side
of the wall. The earth within the trench is a yellow-tan
sandy infll with small feldstones and differs from the
bedded stream deposits, mostly rounded pebbles and
sand, through which it cuts. The stone placed at the bot-
tom of the wall is considerably larger than the others
and probably acted as a foundation stone for the wall as
it was built into the trench.
The location of the wall falls very close to the route
of A. Kuznetsovs predictive model of the route of the
Isklar aqueduct based on least-cost analysis (see above,
Fig. 6). The preserved wall is probably the foundation of
Figure 9. East side of the standing pier of Kavakl Dere bridge of Isklar aqueduct, looking southwest
(A008).
250 ANGELA R. COMMITO and FELIPE ROJAS
the pier of an aqueduct bridge. Without rock outcrops
on which to construct the bridge, the builders founded
the bridge pier deep within the streambed deposits of
the east bank, using the trench on the south side as a
workspace, which they flled in after construction. The
narrowness of the wall, which is less than half as wide as
the bridge pier on the Kavakl Dere, suggests that the
conduit was perhaps only 0.4 m or so wide internally
and that here it ran close to the ground, perhaps even in
a pipeline, rather than atop a tall substructure.
Like the Seki aqueduct, the Isklar aqueduct was
probably built in the frst century A.D. to supply new
buildings in Aphrodisias, such as a bath complex, and
to support the needs of a growing population in the val-
ley. The aqueduct may also have supplied water to agri-
cultural felds in the northern portion of the valley.
3. Derince Dere Aqueduct
The two documented structures of the Derince Dere
pipeline suggest that its route overlapped with that of
the Isklar aqueduct and tapped the same abundant
springs somewhere along the western portion of the
Baba Dag range (see Fig. 6). The northern structure is an
open channel along the west (right) bank of the I
ncir
Deresi in the vicinity of the rock-cut tunnel of the Isklar
aqueduct discussed above (E054). It can be traced for ap-
proximately 280 m. The channel rests on a substructure
of parallel walls, each ranging from 0.30 m to 0.60 m
thick, built of roughly worked schist slabs bonded with
a tan-white lime mortar with densely packed stone in-
clusions smaller than 0.005 m. In several areas, the wall
closer to the streambed has fallen away, indicating that
the two walls are not bonded together. The channel runs
along the west bank of the stream but maintains a
roughly level elevation, while the streambed rapidly
loses elevation as it travels south. Therefore, the channel
begins at almost the same level as the streambed, but as
it travels south, it gets higher and higher above the
stream below.
Where the channel level is close to the level of the
streambed, the channel foor is visible. The channel is
0.33 m wide and has a bedding of pink mortar, 0.02 m
thick, covering the channel foor and preserved to a
height of 0.03 m along the interior walls. The pink mor-
tar is laid directly on top of the masonry substructure.
Layered calcium carbonate deposits, 0.01 m thick in
total, cover all preserved portions of the pink mortar,
suggesting that water fowed in the open channel rather
than in a sealed pipeline laid on its foor.
The channel is located at a considerably higher ele-
vation than the rock-cut tunnel of the Isklar aqueduct,
a second indication, in addition to the style of masonry,
that it is not part of the same system. The original pink
mortar channel was refashioned in more recent times by
the application of hard, white cement to form an altered,
V-shaped conduit. A local informant said the conduit
was reworked with cement recently and was used to
carry water all the way south to the town of Isklar
more than 2 km away.
The pink mortar channel could be the remains of a
later system that used the same water source as the
Roman Isklar aqueduct. The relatively small size of the
channel, the lack of a mortared rubble core within its
substructure, and the survival and recent reuse of the
channel make it signifcantly different from the remains
of the more substantial Isklar aqueduct. Even though
the pattern of calcium carbonate deposits suggests that
at some point water ran in the open channel rather than
in pipes, we have decided to group this structure with
the remains of an aqueduct located approximately 5 km
downstream on the Derince Dere, which obviously ran
in a pipeline. The two structures appear to be part of a
series of waterworks built after the Roman period to
carry water from the important springs on the Baba
Dag ridge southward to Isklar and Aphrodisias.
The remains on the Derince Dere (a local name for the
Isklar Deresi used throughout this article) consist of the
substructure of a bridge that must have crossed the dere
Figure 10. Masonry wall of Isklar aqueduct on Derince
Dere, looking southeast (E050).
THE AQUEDUCTS OF APHRODISIAS 251
and been associated with a terracotta pipeline still pre-
served on the opposite stream bank (E055) (Fig. 11). The
substructure is built of mortared rubble faced with petit
appareil masonry of roughly worked feldstones of
widely varying sizes and materials. It is 1.33 m wide and
preserved to a height of 1.8 m and length of 40 m. The
substructure approaches the west (right) stream bank
from the northwest, but it turns slightly to the east about
4 m from its eastern endpoint, perhaps to cross the stream
at a slightly less acute angle. Almost directly across the
stream from this substructure, at least six segments of a
terracotta pipeline are visible, placed in a trench dug into
the eroding shale of the east (left) bank. Erosion of the
rock has caused the collapse of the side of the trench fac-
ing the streambed, revealing the terracotta pipeline
within. The terracotta pipe segments, joined with white
mortar, are each about 0.5 m in length and 0.15 m in di-
ameter. The pipes have a smooth exterior surface but are
gently ribbed inside, and two parallel grooves circum-
scribe the upstream end of each pipe segment.
The Derince Dere pipeline is the only aqueduct in the
region along which terracotta pipes have been found,
and these pipes match those found in the Ottoman
baths at Aphrodisias. The pipeline may have entered the
city from the east where the remains of what appear to
be two pressure-regulating water towers have been
identifed, as discussed below in Technical Analysis,
Suterazi (Fig. 12). As a likely supplier of water for the
Ottoman settlement, this aqueduct underscores the
longevity of human occupation at the site of Aphro-
disias and the continued importance of the springs on
the western slopes of the Baba Dag range in supporting
these communities.
Water from the Lower Valley
The remains of two aqueducts located west of Aphro-
disias help to form a more complete picture of water
management in the Morsynus river valley. Because both
are open-conduit aqueducts run by gravity and are situ-
ated at levels more than 50 m lower in elevation than the
city, they could not have supplied Aphrodisias. Instead,
they may have carried water to a settlement, farmstead,
or villa, or provided water for the irrigation of sur-
rounding agricultural felds. Combined with evidence
for farmsteads, settlements, and agricultural blocks in
this area of the valley just west of Aphrodisias, the iden-
tifcation of the Kavakl Dere and ren Deresi aque-
ducts provides evidence for a valley dense with human
occupation that depended on locally available natural re-
sources such as water.
The remains of these two aqueducts have been iden-
tifed at only a single location each, so their respective
routes and destinations must be reconstructed based on
other factors. Because the western part of the Morsynus
river valley receives the runoff of the springs and surface
drainage of the Baba Dag range, it has deep, narrow
streams that provide a more abundant water supply
than in the central part of the valley. These aqueducts
may have tapped springs along the Baba Dag range or
the upper reaches of these relatively high-volume
streams. While it is possible that they supplied agricul-
tural irrigation systems, their location in a relatively
well-watered part of the valley suggests that they were
constructed to bring water from a higher elevation to a
destination at a lower elevation that would not other-
wise have received water of adequate volume or qual-
ity. Likely candidates for this type of water requirement
are the settlements, farmsteads, or villas located in the
area at elevations lower than the documented remains.
Therefore, the two aqueducts in the lower valley were
probably built in association with one or more of these
sites dated to the Roman or Byzantine periods.
4. Kavakl Dere Aqueduct
The conduit of the Kavakl Dere aqueduct is visible in
cross section in the west (right) bank of the Kavakl
Dere almost 5 km west of Aphrodisias (E053) (Fig. 13).
The water of the meandering Kavakl Dere must have
facilitated the collapse of the subterranean conduit at
this point where it ran close to the stream. The vaulted
conduit is built of mortared rubble masonry in a single
layer of feldstones including schist and quartz and
smoother stream stones, which are between 0.20 m and
0.30 m long, with smaller stones of 0.05 to 0.10 m in-
terspersed throughout. The masonry is bonded with a
blue-gray lime mortar with densely packed and very
small stone inclusions about 0.001 to 0.002 m in size.
Interesting to note, the foor of the conduit is not con-
structed of the same rubble masonry. Only a sand-rich
sediment layer remains of the conduit foor, which must
have originally been equipped with a more substantial
bedding, perhaps a hydraulic mortar, that has disinte-
grated. The interior dimensions of the conduit are 0.47
Figure 11. Bridge substructure and pipeline on Derince Dere,
looking southeast (E055), with arrow indicating remains of
trench in east (left) bank into which pipeline was placed.
252 ANGELA R. COMMITO and FELIPE ROJAS
to 0.50 m wide and 0.77 m high. The thickness of the
walls and vault ranges between 0.20 m and 0.30 m.
It is diffcult to reconstruct the relationship between
the streambed and the conduit when it was originally
built. The conduit is situated about 3 m above the cur-
rent streambed and 1.8 m below the ground level of the
bank. A construction trench above the conduit extends
all the way to the current ground level and cuts through
multiple layers of stream deposits. After the bedding
and vaulted masonry conduit were constructed within
the trench, mortared rubble was piled on top of the
vault, so that the exterior of the conduit appears rectan-
gular in section. The trench was then backflled.
The conduit is almost completely flled with the same
red, iron-rich sediment as that of the stream bank sur-
rounding it. The sediment in the conduit is about 0.64 m
high, the same height as the calcareous deposits lining
the north interior wall, and includes no stones larger
than 0.05 m. Presumably the fll is sediment deposited in
the conduit when water was still fowing through it. If
so, the aqueduct may have tapped water from a sedi-
ment-laden stream almost identical to that of the
Kavakl Dere, rather than from one of the springs on
the Baba Dag range. In any case, slow water fow al-
lowed both the accumulation of calcium carbonate and
the deposition of suspended sediment.
5. ren Deresi Aqueduct
As with the Kavakl Dere aqueduct, the vaulted and
mortared masonry conduit of the ren Deresi aqueduct
is visible only where it has been exposed in the west
(right) bank of the ren Deresi, where the stream runs
just over 3 km west of Aphrodisias (A029) (Fig. 14). In
the segment currently visible, the conduit wall closer to
the stream has collapsed, revealing both the longitudinal
section and cross section of the structure. Also pre-
served is a rectangular shaft, possibly for maintenance
and inspection.
In cross section, the interior of the conduit is 0.50 m
wide with a 0.30-m wide channel centered in the foor.
The channel is 0.90 m in height, and the rest of the con-
duit rises another 0.90 m above the channel. The con-
duit sidewalls are 0.400.50 m thick. The vault is made
of a single layer of feldstones with rubble packing on
the exterior, all bonded with a beige lime mortar with
small stone inclusions. However, the exposed interior
surface of the conduit wall reveals a different treatment.
A very hard mortar with a reddish orange color covers
not the stones of the wall but their interstices. This
mortar could simply be the same mortar as that used in
the vault, stained by iron-rich sediment infll, or it
could be the remains of a waterproofng mortar now
so damaged that it is not easily distinguishable as an
even coating.
Figure 12. One of two possible pressure-regulating water
towers (suterazi) at Aphrodisias, outside the city to the
northeast.
Figure 13. Conduit of Kavakl Dere aqueduct, visible in
cross section in west (right) bank of Kavakl Dere (E053)
(1:100).
THE AQUEDUCTS OF APHRODISIAS 253
An opening in the vaults roof leads to an inspection
shaft that rises vertically from the conduit and is pre-
served to a height of 0.3 m above the vault. Both the
opening in the vault and the shaft itself have interior
width dimensions of about 0.9 m parallel to the conduit
and 0.5 m perpendicular to the conduit. The well-pre-
served west wall of the shaft, which is built into the
stream bank, is 0.65 m thick. The east walls of the shaft
and conduit have collapsed. The shaft walls perpendi-
cular to the conduit have damaged interior surfaces and
a preserved thickness of 0.4 m. A single putlog hole is
visible in the interior west wall of the shaft and was
likely used to facilitate both construction of the shaft
and access into the conduit after construction.
The conduit is flled with sediment to a height of 1.2
m. The Kavakl Dere and ren Deresi aqueducts were
subterranean conduits that ran adjacent to deep streams
for at least part of their route. Both seem to have even-
tually collapsed as a result of their proximity to these
meandering streams.
Water from Beyond the Valley
6. Timeles Aqueduct
The Timeles aqueduct is the largest and most substantial
of the six identifed aqueducts (Figs. 15, 16). It tapped
tributaries of the Yenidere ay in the Tavas plain and
carried water westward through the hills that mark the
boundary between it and the adjacent valley of the
Morsynus river. This structure, over 25 km in length,
almost certainly corresponds to the Hadrianic-period
aqueduct known from inscriptions of the second quar-
ter of the second century A.D. found at Aphrodisias,
which are discussed at length in Historical Analysis
below.
History of investigation. Remains of the Timeles
aqueduct were frst visited by the survey team in 2006,
when a local informant showed us the standing piers of
a large collapsed aqueduct bridge on the S aban Dere
near the village of Denizoluk and told us that a number
of other such remains could be found in the area (C048)
(see below, Figs. 31, 32). Investigation of those remains
was a major focus of the 2007 season. Our objectives
were to determine whether all the remains belonged to
the same aqueduct, and then to trace that aqueduct as
completely as possible from its source to its ultimate
destination, while at the same time documenting all
major bridges, conduits, tunnels, and shafts with digital
photographs and architectural section and elevation
drawings.
We began our investigation in 2007 by visiting struc-
tures known to our local informants, who frst showed
us a series of tunnels made visible by a recent road cut
along the northern banks of the Kepiz ay, near the vil-
lage of Gzelky on the Tavas plain (G002). We met with
residents of the villages of Kayapnar and Yesilky to ex-
amine the shafts, tunnels, and bridges known to them.
Exploration over two days of the areas between these
known points led to the discovery of further remains.
During the fnal season of investigation of the aque-
ducts in 2008, we focused on completing the architec-
tural records of the surviving bridges, conduits, and
shafts. We were also shown two additional bridges in
the foor of the Morsynus river valley, which provide
crucial information about the path of the aqueduct be-
tween the ridge separating the Morsynus and Timeles
valleys and Aphrodisias (see Fig 15, E058 and E059).
The source and Kepiz ay. Using the methods out-
lined above, we were able to trace the path of the aque-
duct and reconstruct its route from the presumed source
to Aphrodisias. We identifed an exposed segment of the
mortared masonry walls of the conduit as the eastern-
most trace of the aqueduct, following the contours of
Figure 14. Conduit of ren Deresi aqueduct, visible in the west (right) bank of ren Deresi (A029)
(1:100).
F
i
g
u
r
e
1
5
.
R
e
c
o
n
s
t
r
u
c
t
e
d
p
a
t
h
o
f
T
i
m
e
l
e
s
a
q
u
e
d
u
c
t
b
a
s
e
d
o
n
s
t
a
n
d
i
n
g
r
e
m
a
i
n
s
a
n
d
l
e
a
s
t
-
c
o
s
t
a
n
a
l
y
s
i
s
p
r
e
d
i
c
t
i
v
e
m
o
d
e
l
i
n
g
(
1
:
1
0
0
,
0
0
0
)
.
A
m
a
p
o
f
t
h
e
o
u
t
l
i
n
e
d
f
e
l
d
i
s
p
r
o
v
i
d
e
d
i
n
F
i
g
u
r
e
1
6
.
F
i
g
u
r
e
1
6
.
E
a
s
t
e
r
n
(
u
p
s
t
r
e
a
m
)
p
o
r
t
i
o
n
o
f
r
e
c
o
n
s
t
r
u
c
t
e
d
p
a
t
h
o
f
T
i
m
e
l
e
s
a
q
u
e
d
u
c
t
,
f
r
o
m
K
e
p
i
z
a
y
t
o
K
a
y
a
p
n
a
r
,
b
a
s
e
d
o
n
s
t
a
n
d
i
n
g
r
e
m
a
i
n
s
a
n
d
l
e
a
s
t
-
c
o
s
t
a
n
a
l
y
s
i
s
p
r
e
d
i
c
t
i
v
e
m
o
d
e
l
i
n
g
(
1
:
5
0
,
0
0
0
)
.
256 ANGELA R. COMMITO and FELIPE ROJAS
the north (right) bank of the Kepiz ay near the mod-
ern village of Gzelky (E032) (see Fig. 16). Originally
running underground, the conduit has been exposed by
erosion of the bank. A modern dam has been built near
the exposed conduit at the point where several small
tributaries join together to form the Kepiz ay, which
itself feeds into the Timeles river (Yenidere ay) as it
skirts the western edge of the Tavas plain.
The construction of the modern dam has possibily
destroyed or obscured the beginning of the aqueduct.
The area around the dam is certainly a viable site for
the aqueducts source. According to local informants,
the Kepiz ay and its headwaters used to be more co-
pious and clean than they are today, and until quite re-
cently the Kepiz ay fed canals that powered a series
of Ottoman mills located downstream of the dam and
just east of alky, one of which remains standing.
28
However, the source of the aqueduct could have been
as far away as Krkpnar (Forty Springs), an area lo-
cated about 5 km east of alky noted by Robert and
Robert in their description of the Tavas plain as partic-
ularly well watered.
29
Today Krkpnar is an important
source of water for the residents of the Tavas plain. The
exact source of the Timeles aqueduct therefore remains
unknown.
Starting from the easternmost remains noted above,
the conduit continued to hug the north bank of the
Kepiz ay and then passed over a small tributary dere
atop a bridge that has been refashioned as part of a later
water-carrying system, apparently connected to the Ot-
toman mills mentioned above (E015) (Fig. 17). Both
piers remain standing, but only the lower nine courses
can be securely identifed as part of the Roman aque-
duct bridge. The bridge is constructed of mortared rub-
ble faced with petit appareil masonry of roughly
worked blocks of schist and quartz, leveled with chink-
ing stones and bonded with a beige lime mortar with
stone inclusions smaller than 0.02 m. The bridge sur-
vives to a height of 3 m and is 3.20 m wide. The total
length of the bridge is about 9 m, and the distance be-
tween the piers is 2.7 m. The presumably Ottoman con-
duit, open on top and coated in white waterproofng
mortar, is 1.51.7 m wide and 0.8 m deep. Additional
masonry, possibly for buttressing the bridge, is joined to
the downstream (south) side of the piers.
After passing over this bridge, the aqueduct conduit
resumed its route along the north bank of the Kepiz
ay. Construction of a modern road along the bank,
following the same path as the aqueduct, has scraped
away the contours of the hillside, revealing sections of
the conduit and two tunnels perpendicular to the con-
duit that provided access from the stream bank to the
conduit itself where it ran deep within the hillside.
These tunnels, dug through sandy conglomerate rock,
have a roughly vaulted profle. The eastern of the two
tunnels is 0.7 m wide at the foor and 0.4 m wide at the
top and can be entered and followed for almost 11 m
into the hillside, at which point it connects to the aque-
ducts masonry conduit (G001). As the tunnel ap-
proaches this intersection point, it gradually increases
in height from 0.8 m at the entry point to about 1 m, as
sediment from wind, erosion, and road construction has
been deposited on the foor of the exposed tunnel. The
tunnel connects directly to the conduit, which is a
vaulted structure built of worked blocks bonded with a
white lime mortar with densely packed stone inclusions.
The blocks, roughly fashioned from the local conglom-
erate outcrops, are 0.300.40 m long and 0.20 m high.
Calcium carbonate deposits (sinter) created by slowly
fowing water cover the interior walls to a height of
about 1 m from the current level of the conduit foor,
which is covered by infll. Putlog holes are located along
the walls at a height of fve courses, or just less than 1 m,
above the current foor level. Located at the spring point
of the vaulted ceiling, these putlog holes held a wooden
centering framework on which the stones of the vault
were laid, mortared, and allowed to set.
The interior of the conduit is about 1.7 m wide and
1.7 m high and can be followed in both directions. To
the east, it curves slightly north to follow the contours
of the dere for 27 m, after which it is impassable. To the
west, it ends after 8.4 m where the modern road cut has
scraped away the stream bank, removing the conduit
along with it. From this point, a longitudinal section of
Figure 17. Bridge on feeder dere of Kepiz ay with
Ottoman and modern reconstructions, looking northeast;
only lower courses are part of original Roman aqueduct
bridge (E015).
28. In this area, along the Kepiz ay east of alky, we doc-
umented a masonry-lined shaft (E057) connected to a rock-cut
tunnel (E056) that pierces a large bedrock outcrop (see Fig. 1).
While this tunnel could be connected to the Roman aqueduct, it
appears instead to have directed water toward a later, perhaps Ot-
toman mill located next to the stream, now defunct and used to
shelter sheep.
29. Robert and Robert, La Carie 2 (supra n. 12) 4647.
THE AQUEDUCTS OF APHRODISIAS 257
the conduit becomes visible along the road cut for about
270 m (G002, E013, E014) (Fig. 18). The regular blocks
that formed the interior of the conduit have been lost,
revealing the mortared rubble packing that supported
both the conduit foor and the northern sidewall. The
mortared rubble packing along the sidewall is 0.7 m
thick and is constructed of roughly regular blocks of the
local conglomerate rock laid in courses of varying
heights. The mortared rubble bedding below the con-
duit foor is 0.2 m thick and covered with 0.03 m of cal-
careous deposits. A second level of calcareous deposits,
0.035 m thick, is located 0.3 m above the lower layer,
with a layer of sediment and pebbly fll in between. This
double layer of sinter, separated by deposits of sediment
and pebbles, could be the result of two distinct periods
of use.
The intersection point between the eastern access tun-
nel and the masonry conduit was left open for at least
some of the time during which the aqueduct was in use.
Calcareous deposits line the conduit blocks and rubble
packing where they adjoin the tunnel, indicating that the
water fowed freely in this open space. The distance
between the current foor levels of the tunnel and the
conduit is 0.4 m, but the calcareous deposits extend to a
height of 1 m above the current foor of the conduit. To
prevent the leakage of water from the conduit through
the access tunnel, there must have been a device to seal
off the conduit. The junction could have been blocked
up by temporary rubble masonry, torn down and rebuilt
as needed during maintenance, or possibly used as an
off-take point.
Aqueduct conduits running through soil close to the
ground surface could be built in relatively shallow
trenches, covered with backfll after construction and
accessed by short, vertical inspection shafts if neces-
sary.
30
The conduit along the Kepiz ay, however, ran
through both earth and rock and at some places ran
deep within the hillside of the stream bank, making con-
struction in a trench impractical. Instead, the builders
pierced the stream bank with horizontal access tunnels
through the rock and then dug at a perpendicular to cre-
ate a tunnel into which they then laid the masonry con-
duit of the aqueduct. These horizontal access tunnels,
used for construction and possibly maintenance as well,
performed the same function as the more typical verti-
cal shafts, either cut through rock or lined with ma-
sonry, found farther down the Timeles aqueduct.
31
After following the north bank of the Kepiz ay,
Figure 18. Longitudinal section of subterranean conduit visible along road cut by Kepiz ay, looking
northeast (G002).
30. As, e.g., on the Eifel aqueduct at Cologne, Germany; see
A. T. Hodge, Roman Aqueducts and Water Supply, 2nd ed. (Lon-
don 2002) 9394.
31. Vertical shafts have been documented along well-known
aqueducts such as those supplying Carthage and Thugga and the
Gier aqueduct of Lyon. Carthage and Thugga: A. I. Wilson,
Water Management and Usage in Roman North Africa: A Social
and Technological Study (D. Phil., University of Oxford 1997)
67. Lyon: J. Burdy, Les aqueducs romains de Lyon (Lyon 2002)
12327. See Technical Analysis, Tunnels and Shafts below for
more information.
258 ANGELA R. COMMITO and FELIPE ROJAS
the aqueduct turned to the northwest and crossed an-
other small feeder dere. The second preserved access
tunnel is located about 350 m west of the frst, on the
other side of the feeder dere, but cannot be entered be-
cause of sediment infll (E003). The local informant who
showed us the access tunnels said that there had once
been a bridge crossing the feeder dere, but no trace of
the bridge remains visible today. Other evidence for the
aqueduct in this area includes a curved fragment of pre-
cipitated calcium carbonate. Presumably formed on the
inside of a terracotta pipe, the fragment may indicate
that the Roman aqueduct conduit was tapped for use in
a nearby farmstead, village, or for irrigation. Alterna-
tively, the pipe could have been associated with the
Ottoman mills in the area. The possibility that the
Timeles aqueduct provided water for local consumption
is in evidence at several points along its path to Aphro-
disias and is discussed in Technical Analysis, Rural
Water Supply, below.
Yahs iler plateau. After crossing this bridge, the
aqueduct conduit entered a tunnel that ran up to 50 m
underground for almost 2 km through a wide, fat
plateau south of the modern village of Yahsiler. Over
this distance, on top of the plateau, mounds of rock
chips at ground level mark the course of the aqueduct
conduit deep below the surface (east to west: E049,
E048, E036). While one mound to the east of this series
has been scattered by agricultural activity, these three
mounds are large features highly visible in the landscape
and even in satellite imagery (Fig. 19). The striking re-
semblance of the mounds, about 10 m in diameter, to
tumulus tombs encouraged an attempted looting of one
of them. The deep looters trench revealed a thick layer
of rock chips, 2.3 m deep, created during construction
of the aqueduct.
These mounds are heaps of backdirt (spoil) created
as construction crews dug deep shafts from the surface
down to the level of the tunnel. Creating vertical shafts
rather than boring straight through the plateau provided
several points for surveying the orientation and gradient
of the tunnel, helped expedite the tunneling process by
opening up multiple work surfaces from which workers
could dig simultaneously, and provided air to the work-
ers and exits for the removal of backdirt.
To the west of the mounds, a deep depression in the
earth marks the location of a ffth shaft, with traces of
masonry lining the entry point (E052). In contrast to
the masonry-lined shafts and large rock-cut shaft far-
ther along the aqueduct, discussed below, there is no
clear evidence that these mounds corresponded to shafts
that were left open after construction. Two additional
shafts can be tentatively identifed using satellite im-
agery, in which the scattered rock chips of the mounds
appear as white blotches distinct from the agricultural
felds surrounding them. The seven possible shafts are
placed at various intervals ranging from 87 m to 550 m.
The three intact mounds are separated by intervals of
87 m and 98 m, representing the shortest intervals iden-
tifed. Shafts could have originally been placed every
90100 m, or the spacing could have been inconsistent.
Figure 19. Satellite image showing three mounds of rock chips and one masonry-lined shaft on the
Yahsiler plateau (right to left), corresponding to the course of the aqueduct conduit underground, and
remains of the bridge at far left where the aqueduct crossed Ciglek Deresi (east to west: E049, E048,
E036, E052, E033). Google Earth satellite image ( 2010 Google, 2011 DigitalGlobe).
THE AQUEDUCTS OF APHRODISIAS 259
Our local informant said that earth piles from the dig-
ging of shafts or inspection points used to be discernible
every 100 m or so. A more detailed consideration of
these shafts and the tunnel to which they connect is pro-
vided in Technical Analysis, Tunnels and Shafts,
below.
Ciglek Deresi. The aqueduct emerged from its sub-
terranean tunnel under the Yahsiler plateau to cross two
closely spaced bridges, frst turning north to pass over
a small gully, then turning west along a ridge to cross
the larger Cig lek Deresi (Figs. 2022). The frst bridge,
which must have spanned about 8 m to cross the gully,
no longer stands, but traces of the masonry conduit on
either side of the gully indicate its original length
(E034). On the south side of the gully, the conduit is
well preserved and can be entered where it runs under-
ground (E035). On the north side of the gully, traces of
the conduit sidewalls connect to the large bridge that
crosses the Cig lek Deresi itself (E033).
Where the conduit on the south side of the gully con-
nected to the aqueduct bridge that no longer stands, a
clear cross section of its foundation, foor, and the lower
portion of its sidewalls is visible, embedded in the
eroded hillside (E035) (Fig. 23). Here the conduit is 1.50
m wide, with sidewalls 1.15 m (east) and 1.18 m (west)
thick. The conduit sidewalls are built of regular blocks
of the local sandy conglomerate bonded with mortar
and topped with a barrel vault. These blocks are 0.10
0.40 m long, 0.160.20 m high, and 0.150.20 m deep.
Around these blocks, mortared rubble packing consist-
ing of stones varying greatly in size and shape supported
the vaulted conduit. The sidewalls are bonded with
pink-beige lime mortar with densely packed stone in-
clusions smaller than 0.01 m.
The sidewalls and conduit foor rest upon a founda-
tion layer of smaller stones 0.300.35 m thick. Above
this rubble foundation, the conduit foor is equipped
with a bedding of white-gray lime mortar with densely
packed stone inclusions smaller than 0.01 m. The mor-
tar bedding is 0.25 m thick under the center of the con-
duit foor but tapers to a thickness of 0.10 m about
0.200.35 m from both edges of the foor. On top of this
white-gray mortar bedding is a very thin layer, 0.001 m
thick, of pink waterproofng mortar, created by the use
of crushed brick or the local, iron-rich, sandy con-
glomerates, as used in the mortar of the conduit side-
walls. On top of the pink waterproofng mortar is a
layer of sinter, 0.05 m thick, created by the gradual ac-
cumulation of precipitated calcium carbonate in layers,
each less than 0.001 m thick.
The conduit arrived from underneath the Yahsiler
plateau to the east, continued running west within a
ridge between two dry gullies perpendicular to the
broad Cig lek Deresi, and turned north at a 125 angle to
cross a dry gully over the bridge that no longer remains
(E034). Following the conduit from the cross section
visible in the south side of the dry gully and moving up
the course of the aqueduct, the conduit has collapsed
and is flled with earth and vegetation for 6.7 m. In this
segment, only the lower few courses of the conduit side-
Figure 20. Where conduit emerges from hillside in Ciglek Deresi (left), a cross section is visible (marked
with arrow, looking southeast) (E035). Piers of the large bridge are visible at right (E033).
F
i
g
u
r
e
2
1
.
B
r
i
d
g
e
s
a
n
d
t
u
n
n
e
l
i
n
C
i
g
l
e
k
D
e
r
e
s
i
(
E
0
3
3
,
E
0
3
4
,
E
0
3
5
)
(
1
:
4
0
0
)
.
F
i
g
u
r
e
2
2
.
P
l
a
n
o
f
b
r
i
d
g
e
s
a
n
d
t
u
n
n
e
l
i
n
C
i
g
l
e
k
D
e
r
e
s
i
(
E
0
3
3
,
E
0
3
4
,
E
0
3
5
)
(
1
:
4
0
0
)
.
262 ANGELA R. COMMITO and FELIPE ROJAS
walls are standing. After this segment, the conduit is
fully preserved within the ridge and can be entered for
about 22 m, at which point the vault has collapsed. A
short segment of conduit can be entered again after this
collapse, but a second collapse seals it off completely.
The conduit is 1.50 m wide, and heavy sinter deposits
begin about 0.95 m down from the top of the rubble
vault. Much sediment and debris cover the conduit
foor. Measuring from the foor as exposed in the cross
section of the conduit visible by the missing bridge, the
original height of the conduit can be calculated at 2.25
m. Assuming that the sinter deposits on the conduit
foor and sidewalls were laid at the same time, the water
could have run 1.30 m deep, allowing for the rough es-
timate of discharge discussed in Technical Analysis,
Discharge, below. At the current foor level of the con-
duit, the sinter is 0.2 m thick in some places.
The masonry of the conduit is not uniform but can
be divided into three distinct segments. As one enters
the conduit, the frst segment extends 8 m on the outer
sidewall (west) of the turn and 7 m on the inner (east),
which is executed more sharply than the gently curving
outer sidewall. The rubble used in the vault consists of
roughly worked stones varying greatly in size and
arranged like voussoirs in an arch. About 3.5 m into the
conduit, on the west sidewall only, there is 3-m stretch
extending to the end of the rubble-vault segment, in
which three courses of small stones are used in place of
one course of regular blocks. This stretch occurs along
the outer (west) sidewall exactly where the conduit
curves. The lime mortar of the rubble vault is beige pink
with small, loosely packed stone inclusions.
Where the conduit turns, a clear seam is visible sepa-
rating the rubble-vault segment from the adjacent seg-
ment in which the vault is constructed of the same
regular blocks as the sidewalls (Fig. 24). These two seg-
ments are different in several ways. First, in the two
other places where the intact aqueduct conduit is visible,
at Kepiz ay (G001) and at Yesilky (E006), both the
sidewalls and the vault are constructed of regular, rec-
tangular blocks, just like this second segment at Cig lek
Deresi. The rubble vault is thus an anomaly. Second, at
the seam where the rubble vault meets the regular vault
at the conduit bend, the two are offset on the outer
(west) sidewall by 0.100.13 m. In addition, a deep pit
dug into the sediment infll of the conduit has revealed
the profle of the west sidewall of the rubble-vault seg-
ment near the point where it meets the regular-vault seg-
ment. The upper courses of the sidewall are set back
0.050.10 m from the lower courses revealed within the
pit. Third, the putlog holes in the rubble-vault segment
are located one course above the putlog holes of the reg-
ular-vault segment. Fourth, the mortar of the rubble
vault seems to have less densely packed stone inclusions
than the mortar used in the sidewalls of the conduit
where it emerged to cross the bridge that no longer
Figure 23. Conduit exposed in hillside and inside tunnel in Cig lek Deresi (E035) (1:100).
THE AQUEDUCTS OF APHRODISIAS 263
stands (E034). Finally, at the entrance to the conduit,
where the vault has collapsed, the east side of the vault is
constructed with the same regular, rectangular blocks
used in the sidewalls and in the regular-vault segment.
These blocks are displaced about 0.08 m lower than the
level of the rubble vault and could be the remainder of
the original conduit, which would explain the displace-
ment as well. Alternatively, they could have been reused
during the construction of the rubble vault and displaced
as a result of the collapse at this point.
Deep inside the hill the conduit was built inside a
rock-cut tunnel. Where it emerged to cross the small
gully, however, it was built in earth, probably in a trench
reflled with backdirt. These two construction methods
may explain the differences in vault construction seen
within the conduit. The builders may have chosen to
construct a rubble vault where the conduit was laid in
earth, relatively close to the ground surface above. Al-
ternatively, the rubble-vault segment may be a later re-
building of the original conduit where it turned, and
therefore was more susceptible to damage by moving
water, and exited the hillside. In this case, the original
conduit would be represented by the second segment
with its vault of regular blocks identical to those of the
sidewalls.
About 8.8 m upstream from the seam and the con-
duit bend, a third segment begins, in which the top por-
tion of the vault of regular, rectangular worked blocks
appears to have been rebuilt in rubble masonry with
thickly applied white-blue lime mortar with large,
densely packed stone inclusions, most between 0.005 m
and 0.012 m. The added rubble masonry is 0.60 m wide
across the top of the vault and has a level profle, mak-
ing it not a true vault but a much weaker structure, no
doubt responsible for its collapse about 5 m farther
down the conduit. The roughly worked stones of the
false vault are placed lengthwise across the width of
the conduit, in contrast to the stones of the rubble-vault
segment discussed above, which are placed like vous-
soirs in an arch. The stones used in the false vault in-
clude both feldstones and roughly worked blocks of
the local conglomeritic sandstone, just like the blocks
used in the sidewalls and in the vault and sidewalls of
the adjacent regular-vault segment, suggesting that per-
haps some of the blocks of the original conduit were
reused to create this false vault. The thickly applied
mortar protrudes below the level of the vault of regular,
rectangular blocks built with a centering framework,
further suggesting that this false vault is a later rebuild-
ing, intended to mend a collapse or patch up the con-
duit after maintenance of some sort. In this segment, the
putlog holes are located one course above the putlog
holes of the regular-vault segment. Patches of a smooth,
white lime mortar without inclusions surround the
blocks of the upper sidewalls and lower vault in this seg-
ment as well. The mortar was applied horizontally in
layers, leaving a distinct ribbed pattern.
The false vault was certainly a repair job. Sinter ac-
cumulation up to 0.20 thick on top of the smooth, white
mortar indicates that the aqueduct was still in use after
its application. In one place, a 1-m stretch of sinter has
been intentionally chipped off with a pick, perhaps for
reuse as building material elsewhere.
As noted above, the putlog holes are located at dif-
ferent heights within the three segments, the most ob-
vious difference being their location one course lower in
the regular-vault segment than in the other two seg-
ments. The putlog holes of the rubble-vault segment are
set at intervals ranging from 0.35 to 1.40 m in no clear
pattern and are matched only roughly in pairs across the
conduit. The putlog holes are set at shorter intervals at
the inner part of the conduit curve. In the regular-vault
segment, the putlog holes are set at intervals ranging be-
tween 0.40 m and 2.90 m, and are not matched across
the conduit due to the bend. In the false-vault segment,
the putlog holes are matched in pairs across the conduit,
Figure 24. Interior of conduit before it emerges at Cig lek
Deresi, looking southwest (E035). Where conduit turns, a
seam separates upstream segment vaulted with rectangular
blocks from downstream segment with rubble vault.
F
i
g
u
r
e
2
5
.
T
w
o
-
t
i
e
r
e
d
H
a
r
a
m
i
D
e
r
e
b
r
i
d
g
e
n
e
a
r
Y
e
s
i
l
k
y
(
E
0
0
8
)
(
1
:
4
0
0
)
.
THE AQUEDUCTS OF APHRODISIAS 265
but those on the northeast sidewall are located about
0.14 m below those on the opposite sidewall. The irreg-
ular intervals of the putlog holes suggest rebuilding or
building with purposefully distinct segments, perhaps
to accommodate the bend. Additional discussion of the
technical aspects of conduit construction follows in
Technical Analysis, Conduit, below.
Two piers of an aqueduct bridge are preserved on ei-
ther side of the dry Cig lek Deresi (E033) (see Fig. 21).
The total length of the bridge is approximately 42 m,
and the estimated height of the aqueduct conduit above
the streambed as it ran across the bridge is 11.5 m. The
arch and upper portions of both piers have collapsed,
leaving a scatter of mortared rubble across both sides of
the dere. The piers are constructed of mortared rubble
faced with petit appareil masonry. The facing blocks are
worked from the well-cemented, sandy conglomerates
located throughout the area. Other blocks are fashioned
from locally available low-grade, gray-blue marble and
quartz. The facing blocks and rubble core are bonded
with pink-beige lime mortar with stone inclusions
smaller than 0.01 m. The mortared rubble core is struc-
tured in layers, and the facing blocks are leveled with
small chinking stones.
The base of the west pier is 5.2 m wide and is founded
on a rock outcrop. At eighteen courses (3.6 m) above
the base of the west pier is a leveling course of blocks
larger than the other facing stones, all fashioned with a
chisel from locally available low-grade, gray-blue mar-
ble. The leveling course does not project from the sides
of the pier and appears to be too low to have served as
a spring point for the arch, which would have spanned
just over 6 m.
The west pier is better preserved than the east and has
many in situ facing blocks on the south face, and only a
few on its west and north faces. On both piers, the
downstream (south) side is much better preserved than
the upstream (north) side, which has suffered greater
damage from the forces of wind and weather, as seen on
all the other aqueduct bridges documented. See Tech-
nical Analysis, Bridges, below, on the causes of up-
stream damage.
Yes ilky. From the Cig lek Deresi, the conduit dis-
appeared underground to run northwest for about 1 km
until it once again emerged on a series of bridges cross-
ing deep gorges near the village of Yesilky. The bridges
are constructed of mortared rubble set in layers and
faced with petit appareil masonry. Incorporated into the
facing of the piers are periodic leveling courses, which
project slightly from the masonry above and below
them. These projecting courses are invariably built of
marble, also used for some of the voussoirs, and they
include occasional reused blocks.
The aqueduct conduit frst ran across a tall bridge
above the Harami Dere (E008) (Fig. 25). Although the
bridge is poorly preserved, analysis of the standing re-
mains on both banks of the gorge suggests that the
bridge was originally two-tiered, with a lower arch sup-
porting possibly four arches above, topped by the con-
duit. Traces of the conduit survive on the south (left)
bank of the gorge, indicating that the bridge as a whole
was about 65 m long and 27 m high.
While nothing remains on the north (right) bank of
the gorge except weathered mortared rubble and scat-
tered facing blocks, the bridge is better preserved on the
opposite bank. The few facing blocks of the bridge that
remain in situ are located on the west (downstream) side
of the south pier and are fashioned from the sandy con-
glomeratic outcrops forming the banks of the gorge.
Three kinds of mortar are visible in the remaining sec-
tions of the structure. Ubiquitous is a white lime mor-
tar with stone inclusions no larger than 0.005 m. Also
visible around the crumbling mass of the upper north
pier is a light gray lime mortar with larger stone inclu-
sions up to 0.01 m in size. A third type of lime mortar
can be found on the buttressed middle mass on the
south bank; it has a very loose, pinkish tan matrix with
stone inclusions as large as 0.02 m.
The single arch of the lower tier springs from the
bedrock of the steep south (left) bank of the gorge at a
height of 10 m above the bottom of the gorge. The tier
is preserved to a height of 5.5 m, and the arch originally
spanned a distance of approximately 13 m. The arch,
only 12 courses of which remain standing, is con-
structed of two layers of voussoirs, the upper blocks
carved from the local well-cemented, sandy conglom-
erates and the lower of marble. The maximum preserved
width of the lower arch is about 3 m, though the facing
blocks of the upstream (east) side are missing. Marble
voussoirs and a large mass of mortared rubble litter the
streambed below the arch and farther downstream
along the gorge for at least 20 m. The marble blocks are
variously dressed with anathyrosis, clamp cuttings, lift-
ing bosses, and moldings, indicating that they were re-
cycled as voussoirs.
The second tier of the bridge survives mostly in
traces on the south (left) bank of the gorge and begins
approximately 4.5 m above the remains of the lower tier.
It has been reconstructed as a series of four arches, cen-
tered on the lower arch and each spanning 5.75 m. A
notable feature of the remains is a buttressing structure,
perhaps added after the time of initial construction,
which joins the downstream (west) side of the second
tier in a clear seam. As the upstream (east) side of the
second tier on both banks is heavily damaged and has
no preserved facing blocks, the buttress on the down-
stream side must have been built to offset the destabi-
lizing effects of wind and weather on the tall structure.
Notably, only the second tier, which had to span a much
wider part of the gorge than the tier below, was
266 ANGELA R. COMMITO and FELIPE ROJAS
equipped with a buttress. Approximately 1 m above the
upper tier, the sidewalls of the conduit are visible emerg-
ing from the hillside.
After running underground for 190 m, the aqueduct
then passed over another bridge spanning the Krkm
Deresi on an arch that remains intact and can still be
crossed on foot (E005) (Figs. 26, 27). On the west (right)
bank, a section of the vaulted conduit and tunnel is ex-
posed. Pickmarks line the exposed surfaces of the tunnel,
excavated from the conglomerate bedrock of the gorge,
into which the vaulted conduit was laid using small
blocks of the same rock. The conduit ran approximately
36 m across the bridge and 16.7 m above the gorge below.
Based on the remains of the conduit and tunnel, the
height of the entire structure is reconstructed as 19 m.
Both piers are founded on rock outcrops. The base
of the east pier sits 3.7 m above the base of the west pier.
The base of the west pier is constructed of marble
blocks several times larger than the other facing blocks.
As on the Harami Dere bridge, the upstream side of this
bridge has suffered considerably more damage than the
downstream side, on which facing blocks are preserved.
Nevertheless, the width of the better-preserved west
pier can be reconstructed at 4.3 m. The facing blocks are
Figure 27. North side of Krkm Deresi bridge near Yesilky, looking southwest (E005).
Figure 26. Elevation and section of Krkm Deresi bridge near Yesilky (E005) (1:400).
THE AQUEDUCTS OF APHRODISIAS 267
fashioned from the same sandy conglomerate outcrops
ubiquitous in the region and used for the conduit and
other bridges. The lime mortar is white with small stone
inclusions, most no larger than 0.05 m.
Both piers terminate in projecting courses of marble
blocks, from which the arch of the bridge springs. In-
teresting to note, the top of the projecting course on the
east pier is approximately 0.15 m lower in elevation than
the corresponding course on the west pier, which has an
additional projecting course located approximately
halfway between the upper projecting course and the
base of the pier. The lower projecting course of the west
pier, 2.6 m above the base of the structure, is made of
marble blocks more uniform in size and shape than
those of the upper projecting course, located 3.2 m
above the bottom of the lower course. The arch spans
9.8 m across the gorge and has two layers of voussoirs,
the upper layer being larger than the lower.
After crossing this bridge, the aqueduct turned
southwest, running underground for a short distance
and then emerging in a poorly preserved bridge, of
which only the large footing blocks of sandy conglom-
erate and the weathered mortared rubble core remain
(E007). The east pier of the bridge is 4.4 m wide, with a
maximum preserved height of 6 m. The masonry is
bonded with white lime mortar with stone inclusions,
most smaller than 0.05 m.
Between this bridge and the next one, the conduit ran
underground in a tunnel that can be entered from the
downstream side and followed for more than 150 m
(E006) (Figs. 28, 29). Comparable to the conduit seg-
ments at Kepiz ay (G001) and Cig lek Deresi (E035),
the vaulted conduit is 1.44 m wide and at least 1.9 m
high and is built of regular, rectangular blocks cut from
the local sandy conglomerate bedrock. The conduit
blocks are 0.300.40 m long and 0.20 m high. The
smaller blocks of the leveling course at the spring point
of the vault are 0.30 m long and 0.13 m high. The lime
mortar is beige-pink with densely packed stone inclu-
sions smaller than 0.01 m. No trace of waterproofng
cement or other treatment of the interior surfaces is vis-
ible. No clear sinter level is discernible, and calcareous
deposits lining the vault and walls appear to be the re-
sult of water percolating through the tunnel and con-
duit walls.
Figure 28. Elevation of masonry conduit inside
tunnel near Yesilky (E006) (1:100).
Figure 29. Masonry conduit inside tunnel near Yesilky, looking northeast (E006).
The space between the vaulted conduit and the tun-
nel, excavated from the rock with a head-and-
shoulders profle, is flled with mud and rubble
packing. At 86 m from the entry point, the direction of
the tunnel bends slightly to the north, and approxi-
mately 5 m later returns to its original orientation, per-
haps indicating the meeting point of two construction
crews working in opposite directions. Regular putlog
holes in the segment downstream from the bend placed
at alternating intervals of 0.35 m and 1.40 m indicate that
the centering framework was about 1.75 m long. In con-
trast, the putlog holes in the segment upstream from the
bend appear to be differently spaced, with intervals of
0.75 m and 1.30 m, suggesting the use of centering 2.05
m long.
The conduit emerged from this tunnel to cross a low,
dry dere. A severely eroded mass of mortared rubble on
the south side of the dere is all that remains of the bridge
that would have carried the conduit about 9 m above
the streambed below (G003).
After the series of bridges around Yesilky, the
aqueduct returned underground for 1.5 km before
crossing a narrow gorge on small bridge (E026). Be-
tween these bridges, the top of a masonry-lined shaft
is visible nestled in the contours of a dere hillside
(E051). Now flled in with earth and fallen blocks, the
shaft is 1.25 m by 1.30 m in internal dimension and is
constructed of feldstones of widely varying shape and
size, about 0.20 to 0.50 m long and 0.05 to 0.30 m high.
The stones are bonded with white lime mortar with
small stone inclusions. This shaft was presumably built
to facilitate construction of the subterranean tunnel and
conduit, which lay at least 25 m below ground level. It
may have been backflled after construction, in which
case settling of the earth has revealed the masonry lin-
ing needed to form the shaft until the diggers struck
rock below, as seen around one of the shafts on the
Yahsiler plateau mentioned above (E052) and those
near Denizoluk, discussed below. Alternatively, this
shaft may have remained open as an inspection shaft to
allow access for maintenance. If so, it could have been
flled in after the abandonment of the aqueduct either as
an intentional safety precaution or as the gradual result
of neglect.
The next small bridge, which crosses the Sog anlk
Deresi, is heavily encrusted with calcareous deposits
(E026) (Fig. 30). These deposits were formed by the slow
fow of water leaking from the aqueduct conduit while
water still passed through it. Such extensive calcareous
incrustations could indicate that here the aqueduct was
tapped for local irrigation or some industrial use, per-
haps once the aqueduct no longer supplied the city.
32
Traces of the conduit vault are visible on both sides of
the bridge, indicating that the structure was 5.5 m high
and approximately 20 m long. The dimensions of the
conduit sidewalls, each 1.2 m thick, and foor, 1.50 m
wide, suggest that the whole structure was 3.9 m wide
and therefore retained the same dimensions as the ob-
served segments of conduit located upstream. The up-
stream (north) side of the bridge has deteriorated to the
extent that it is only half as wide as it was originally.
Like the other bridges, this small bridge is built of
mortared rubble faced with petit appareil masonry
bonded with a pinkish tan lime mortar with stone in-
clusions, most less than 0.01 m. The facing blocks are
carved from the local, low-grade, blue-gray marble and
feldstones such as schist. The arch, which springs from
the banks of the gorge, has two layers of voussoirs
carved from the blue-gray marble and spans 4 m across
the gorge.
Denizoluk. Returning underground, the aqueduct
conduit changed direction, running west-northwest
268 ANGELA R. COMMITO and FELIPE ROJAS
32. As seen, e.g., along sections of the Nmes aqueduct and
the frst Arles aqueduct that, no longer suitable to supply the city
of Arles, was restructured to power the mills of Barbegal. Ph. Le-
veau, Research on Roman Aqueducts in the Past Ten Years, in
A. T. Hodge (ed.), Future Currents in Aqueduct Studies (Leeds
1991) 14962, at 15253.
Figure 30. Soganlk Deresi bridge between Yesilky and Denizoluk (E026) (1:200).
along the northern bank of a streambed. Traces of the
conduit sidewalls are visible along the sloping north
bank of the I
nce
Dere, is lined with roughly worked, mortared feld-
stones (C049) (Figs. 33, 34). It is topped by a barrel vault
oriented perpendicular to the streambed and to the
aqueduct conduit underground and built of stones of
the same size. The bottom of the shaft at its point of in-
tersection with the conduit is accessible via a horizon-
tal access conduit, which emerges from the north (left)
bank of the stream. The streambed and the conduit run
parallel to each other, with the access conduit perpendi-
cular to both. The foor of the access conduit lies ap-
proximately 0.1 m lower than the top of the vault of the
aqueduct conduit. The aqueduct conduit itself is cur-
rently flled almost to the top of its vault with mud, de-
bris, and water fowing slowly from east to west.
The structures are currently being used as part of a
modern pumping system, access to which is provided
by a rectangular concrete manhole connected to the an-
cient access conduit. About 2 m above the aqueduct
conduit, the shaft has been sealed with concrete to pre-
vent debris from falling into the water collected in the
modern system.
The internal dimensions of the shaft are 1.45 m by
1.45 m, and the vertical walls are 0.901 m thick. The
stones of the shaft are 0.200.30 m long, 0.100.15 m
high, and 0.200.30 m deep. Most of the vault has col-
lapsed, but presumably it was pierced by a manhole that
would have been covered when not in use. The depth
of the shaft, from the vault to the top of the conduit
below, is approximately 6.3 m. The horizontal access
conduit is barrel-vaulted and built of mortared masonry,
with regularly shaped, rectangular blocks 0.200.30 m
on a side. The access conduit intersects with the vertical
shaft about 5 m down from the top of the shaft. The
conduit, which at the time of investigation during the
summer was flled with water to a depth of 0.1 m, is 2.3
m long, 1.45 m wide, and about 1.25 m high. A pair of
putlog holes under the springing of the vault is located
0.2 m from where the conduit meets the shaft.
At the southern end of the horizontal access conduit,
where it approaches the stream, a barrel-vaulted struc-
ture has been attached, which appears to be a later ad-
dition. A vertical wall approximately 1 m high and built
of stones with their long sides placed vertically, rather
than horizontally as seen on the inspection shaft, con-
nects the two barrel vaults. This barrel-vaulted struc-
ture is more than 1 m long and approximately 1.2 m
wide. The walls below the vault no longer stand, mak-
ing the full height of the structure unknown. The stones
used in this vault are larger than those used in the vault
of the inspection shaft and are at a slight upward angle,
and the mortar of this structure is thickly and unevenly
applied, unlike that of the inspection shaft and access
conduit.
While the vertical shaft was likely built to provide a
permanent point of access for inspection and mainte-
nance, the function of the horizontal conduit is less
clear. It is built so that its foor is at the level of the top
of the vaulted aqueduct conduit, probably to prevent it
from being fooded by water carried in the aqueduct.
There is no trace of waterproofng mortar or sinter on
Figure 34. Masonry inspection shaft on I
nce
Dere near Denizoluk (C044) (1:200).
THE AQUEDUCTS OF APHRODISIAS 273
aqueduct. Alternatively, the tunnel system could have
been used to collect water percolating through the hill-
side and to conduct it into the shaft, where it would
have added to the volume of water already fowing
through the aqueduct below. It may instead have been
used as an air barrier to prevent the seepage of water
into the shaft during its construction, helping to keep
the work surface visible and dry.
The hillside above the shaft is covered with a consid-
erable amount of pottery. It is tempting to interpret this
pottery scatter as either the traces of the construction
crew, temporarily camping on-site to speed up con-
struction time, or the remains of a small settlement or
farmstead. Inhabitants of the settlement or farmstead
could have obtained water for drinking and irrigation
from the tunnel system, acting as a drainage gallery, or,
after the construction of the aqueduct, by drawing
water from the conduit itself. In the latter case the same
type of rope and pulley system used during construc-
tion could have been used to draw water from the con-
duit, essentially using the aqueduct as a well.
When viewed together, the impressive depth and
size of the shaft, the elaborate system of associated
tunnels, the multiple cuttings for a rope and pulley,
and the pottery scatter indicate that this site was a
major focus of construction for the aqueduct. The fact
that over 13 m of the shaft is still exposed suggests that
it was never completely filled in and could have possi-
bly remained open to provide access to the conduit
both for maintenance and for water supply to the local
population.
The Morsynus river valley. The next point we were
able to identify lies about 5 km to the west, near the vil-
lage of Kayapnar in the hills marking the southeast
limit of Morsynus river valley. Over this distance the
aqueduct must have traveled underground from the wa-
tershed of the Timeles river into the Morsynus river val-
ley by piercing a relatively low saddle in the chain of
hills that forms the boundary between these two val-
leys. The hills to the southwest (Avdan Dag ) and north-
east (Tepeky) of this saddle rise to elevations above
1,000 masl. The aqueduct builders chose to tunnel
through this saddle because of its low and level elevation
relative to the ridges on either side. At this point, the
backdirt pile or spoil heap of a shaft and a section of
tunnel are visible at the beginning of a large streambed
known as the Koca Dere (E010).
However, one structure likely related to the aqueduct
was identifed in this area before the aqueduct reached
the Koca Dere. Turning along the contour of a small
hill, a tunnel with a head-and-shoulders profle similar
to the tunnel at Yesilky (E006), but considerably
smaller, may have served as an access tunnel for the sub-
terranean aqueduct conduit (E028). The maximum pre-
served dimensions of the tunnel, cut through rock, are
0.65 m high and 0.85 m wide at the base. The profle
narrows to 0.4 m wide and then 0.2 m at top. The tun-
nel is preserved to a length of 5.5 m.
After tunneling through the low saddle between
Avdan Dag and Tepeky, the aqueduct entered the
Morsynus river valley along the west bank of the Koca
Dere, one of the gorges that drain surface runoff and
spring water on a northbound course from the south-
eastern hills of the valley to the Morsynus river itself.
Near the head of the Koca Dere, the backdirt pile or
spoil heap of a shaft is visible just below the village of
Kayapnar (E010). The Koca Dere soon becomes a nar-
row gorge cutting deep into the southeastern hills of the
valley. About 380 m downstream from the shaft, the
rock of the west (left) bank of the gorge is cut vertically
to create a level bedding for the aqueduct conduit. Pick-
marks are preserved on the vertical rock face (E009)
(Fig. 36). It is unclear whether the rocky stream bank
was originally cut to form a tunnel or an open channel
into which the conduit was laid. While the conduit itself
is not preserved, the local informant who showed us the
remains remembered being able to enter a conduit far-
ther upstream that, according to his description, was
similar to the vaulted masonry conduit documented
elsewhere along the aqueduct.
The narrow Koca Dere broadens as it travels north-
west, merging with other streams until it joins the
Morsynus in the level plain southeast of Aphrodisias.
About 5.7 km downstream from the rock-cut tunnel (or
channel) at Kayapnar, a segment of rubble masonry
conduit is visible in the eroded west (left) bank of the
Koca Dere, several meters above the modern streambed
(E031) (Fig. 37). The conduit is built of unworked feld-
stones of varying sizes, has a fat rather than a vaulted
roof, and is approximately 1 m high and 0.5 m wide in-
ternally, with sidewalls 0.2 m thick. It is located down-
stream from the point where the Timeles aqueduct
presumably turned northeast to head through the
foothills of Koca Dag just northwest of the town of
Figure 36. Vertical cutting with visible pickmarks along Koca
Dere near Kayapnar, looking northwest (E009).
Gkeler. In addition, the conduit in the Koca Dere is
located at least 20 m lower in elevation than the aque-
duct bridges in the Morsynus river valley. These cir-
cumstances, as well as differences in construction,
indicate that this conduit it is not the main conduit of
the original aqueduct. However, its location near the
probable route of the aqueduct suggests that it was con-
nected with it in some way, perhaps as a subsidiary con-
duit that tapped water from the main aqueduct to
supply a nearby village or farmstead or to irrigate felds.
Alternatively, it could be a self-contained, small-scale
aqueduct similar to the Kavakl Dere and ren Deresi
aqueducts discussed above in Isklar-Area Aqueducts.
At the very end of the fnal season of investigation in
2008, a local informant showed us the remains of two
bridges in the Morsynus river valley that spanned streams
feeding into the Morsynus river from the north, and re-
membered being able to enter a tunnel that pierced the
foothills of Koca Dag situated between the Koca Dere
and these streams. It is possible that the aqueduct ran
through this same hill rather than hugging its con-
tourssouth of the two bridges in the river valley before
crossing them and heading west toward Aphrodisias.
The frst of the two bridges crossed by the aqueduct
spanned the Gms (Korkor) Dere (E058) (Fig. 38).
Only one pier remains of this bridge, which is built of
mortared rubble faced with petit appareil masonry and
stands on the north (right) bank of the stream. The length
of the pier has been cut in two to accommodate a dirt
road parallel with the stream. In contrast to all the other
bridges of the aqueduct, the better preserved side is the
upstream (east) side, though in this area of the valley, car-
peted by orchards and agricultural felds, the most dele-
terious force on the bridge was likely not weather but the
reuse of the bridges blocks to build the stone walls that
delineate the felds and even abut the bridge itself.
The pier is preserved to a maximum width of 3.5 m,
length of 12 m, and height of 4.8 m. The rounded feld-
stones that make up the structure are not as internally
consistent in shape and size as those of the other bridges
along the aqueduct, and many of the facing blocks are
larger than those of the other bridges. The lime mortar
is gray with densely packed stone inclusions smaller
than 0.02 m. The pier rests on a rock outcrop forming
the bank of the dere. No leveling courses are incorpo-
rated into the pier, and because nothing remains of the
arch or the conduit that must have originally spanned
the Gms Dere, the height and length of the entire
bridge are unknown. The pier extends to the north into
agricultural felds, where it is no longer traceable.
The second bridge in the Morsynus river valley is the
closest structure to Aphrodisias we were able to identify
(E059) (Fig. 39). The lower courses of both sides of the
arch are well preserved, while the upper portions of the
arch and both piers have been cut down to serve as a
substructure to support a modern concrete channel for
the irrigation of the extensive agricultural felds in the
area. The upstream (north) side has suffered more dam-
age than the downstream side. The bridge, spanning 5.8
m to cross the Seki Deresi, is 5.2 m wide and survives to
a height of 8 m above the streambed below. The visible
274 ANGELA R. COMMITO and FELIPE ROJAS
Figure 37. Cross section of masonry conduit in Koca Dere,
looking northwest (E031).
Figure 38. Bridge on Gms (Korkor) Dere, looking north-
east (E058).
structure is more than 50 m in length and may have
originally been 11 m high.
The bridge is built of mortared rubble faced with petit
appareil masonry. Both piers of the bridge are founded
on the bedrock forming the banks of the Seki Deresi. At
its east and west extremities, the structure is buried under
agricultural felds. The west pier alone extends 34 m from
the streambed before disappearing into the ground. At
about 10 m from the streambed, the west pier is pierced
by a drainage conduit included as part of the original
structure. This conduit must have served, as it does today,
to prevent surface water from gathering along the foun-
dations of the pier and undermining its structural in-
tegrity. Unlike the arches of the other bridges on the
aqueduct, here there is only a single layer of voussoirs,
which incorporate both feldstones and marble blocks.
This structure is especially interesting because it is
the best-preserved above-ground section of the aque-
duct, apart from the bridges. Because the agricultural
felds in which the aqueduct is located have been
worked intensively for centuries since the aqueduct was
abandoned, it is unclear at what point the conduit once
again entered the ground. The apparent absence of other
aqueduct remains west of this point makes it unlikely
that the aqueduct resurfaced before reaching Aphro-
disias itself. Presumably the builders chose to carry the
conduit on a solid substructure across the level valley
because the difference between the elevation to be main-
tained by the conduit and the elevation of the ground
surface was not large enough to necessitate the con-
struction of an arched arcade.
33
No remains of the Time-
les aqueduct have been identifed between this sub-
structure and the city.
TECHNICAL ANALYSIS
Discharge
While the Seki and Isklar aqueducts collected water
from springs in the hills and mountains surrounding the
Morsynus river valley, engineers had to search farther
afeld for a source adequately abundant to supply the
large-scale projects envisioned for the growing city. The
perennial springs along Baba Dag have only a small
recharge area from which they gather snowmelt, and
most of the streams in the valley were likely ephemeral
or intermittent in antiquity as now. Furthermore, the
most abundant perennial springs at Seki and on the
western slopes of the Baba Dag range were already
being exploited to supply the city. The closest area
where plentiful, perennial streams could be found was
in the neighboring Tavas plain, where multiple tribu-
taries run together to form the Timeles river (Yenidere
ay). The Timeles aqueduct stretched southeast from
Aphrodisias to tap these tributaries.
Unfortunately, modern discharge data for these
springs and streams have not been collected, and corre-
sponding ancient data will in any case remain unavail-
able. It should be noted that even perennial streams such
as the Kepiz ay and other tributaries of the Yenidere
ay can have signifcant variability in discharge from
season to season and even from day to day. Analysis of
modern fow rates for the springs that fed Roman aque-
ducts at Metz, Frjus, and Nmes shows that variations
THE AQUEDUCTS OF APHRODISIAS 275
33. Substructures were not uncommon on Roman aqueducts.
E.g., where the conduit of the Gier aqueduct at Lyon ran less than
3 m above ground level, the builders opted to support it with a
solid concrete substructure faced in opus reticulatumrather than
with arches, which were constructed elsewhere along the aque-
duct where the conduit ran at a greater height above the ground.
Figure 39. Elevation and section of Seki Deresi bridge (E059) (1:200).
J. Burdy, Some Directions of Future Research for the Aqueducts
of Lugdunum (Lyon), in Hodge, Future Currents in Aqueduct
Studies (supra n. 32) 2944, at 41; Burdy, Les aqueducs romains de
Lyon (supra n. 31) 5152, fg. 20.
between minimum and maximum daily fow rate, and
between daily discharge of the days within a given
month, were within one to three orders of magnitude.
34
Though it is not clear whether the Timeles aqueduct
tapped the Kepiz ay, another stream, or the springs
feeding these streams, it is important to note that vari-
ability in source output certainly had implications for
source selection, the schedule of aqueduct maintenance,
and the distribution of water within the city itself. Most
important, the fow rate of its source waters determined
the discharge of the aqueduct itself.
Although discharge at the aqueduct source cannot be
measured, an estimate for the potential volume of water
transported by the aqueduct can be tentatively gleaned
by analyzing the accessible segments of conduit. Where
the conduit is exposed in the Cig lek Deresi (E035), it is
1.50 m wide, and heavy sinter deposits begin about 0.95
m down from the top of the rubble vault. Using the
cross section of the conduit made visible by the collapse
of the adjacent bridge, the original height of the conduit
is calculated at 2.25 m. Assuming that sinter deposits on
the conduit foor and sidewalls were laid at the same
time, the water could have run 1.30 m deep. With these
measurements plus values for the average slope of this
section of the aqueduct, the discharge of the aqueduct at
this point can be estimated.
Using both the Bazin and Manning formulae for
mean velocity and two possible values for the Chzy co-
effcient (concerning the roughness of the conduit sur-
face), the result is a minimum potential discharge per day
of 214,963 m
3
and a maximum of 286,589 m
3
(see Ap-
pendix, Calculation 1). Even the lowest estimate for po-
tential daily discharge of the aqueduct214,963 m
3
/24
hoursis extremely high in comparison with estimates
for the aqueducts of Rome. For example, the Aqua Clau-
dia and the Aqua Anio Novus, both built in the frst cen-
tury A.D., had an estimated discharge of 184,220 and
189,520 m
3
/24 hours, respectively.
35
The Anio Novus, in
fact, was the most abundant of Romes eleven aqueducts.
While the mean velocity calculated for the Timeles falls
comfortably within the normal range for Roman
aqueducts, about 1.01.5 m per second, its discharge
does not.
36
The suspect variables, then, are water level
height and gradient. The gradient used for this section
of the aqueduct, from the modern dam to the rock-cut
tunnel at Kayapnar, is 0.95 m/km, certainly not exces-
sively steep in relation to Roman aqueducts in general
or even to other sections of this aqueduct.
More likely, the value of 1.30 m for depth of water at
this point is an overestimation. The accumulations of
sinter on the conduit foor may have been deposited ear-
lier than those on the conduit sidewalls, which instead
represent a later period of water fow after sinter and
sediment accumulation raised the level of the conduit
foor and thus raised the water level as well. The fact that
in this area the conduit was constructed using three dif-
ferent techniques may indicate that the conduit segment
from which the height of sinter on the sidewalls was
taken belongs to a later construction phase not associ-
ated with the original conduit foor, also resulting in an
overestimation of water depth. Unfortunately, the meas-
urements cannot be corrected by comparison to the seg-
ment of conduit visible at Yesilky (E006), because here
no clear maximum sinter level is preserved on the side-
walls. Instead, calcareous deposits lining the vault and
walls appear to be the result of water percolating natu-
rally through the tunnel and conduit masonry. In addi-
tion, the foor level of the conduit visible along the Kepiz
ay (G001) has been signifcantly raised due to erosion
and infll via repeated entry and damage caused by the
construction of the modern road nearby.
As a control, however, the average distance between
the top of the vault and the level of the putlog holes,
which do not contain sinter, in these two sections of con-
duit can provide a rough estimate of maximum water
level. The maximum preserved interior height of the
conduit at Yesilky, less subject to sediment infll than
the conduit at Kepiz ay, is 1.90 m. The average dis-
tance between the top of the vault and the level of the
putlog holes in both segments of conduit is about 0.80 m.
Allowing 0.10 m for sediment accumulation, erosion,
and infll on the conduit foors, we can use 1.00 m as a
tentative estimate of the height of water fowing through
the conduit.
Changing just this one variable has considerable ram-
ifcations for our minimum calculation of the aqueducts
potential daily discharge: the new fgure is 156,989 in-
stead of 214,963 m
3
/24 hours (see Appendix, Calculation
2). In this case, the Bazin formula is used to compute ve-
locity, and a value of 60 is used for the Chzy coeffcient
to obtain the minimum possible velocity. A second po-
tential problem is the gradient, 0.95 m/km, which is cal-
culated over a distance of 11 km, and therefore may not
correspond to the actual gradient of the conduit at this
paticular point. To fesh out the parameters of potential
discharge for the aqueduct, halving the gradient gives an
estimate of almost 111,000 m
3
/24 hours, while doubling
the gradient results in a discharge of about 222,000 m
3
/24
hours (see Appendix, Calculation 3). The aqueduct, then,
likely carried somewhere around 150,000 m
3
of water
into the city each day and, given its remarkably wide
conduit, could have transferred considerably more. The
important point to glean from these calculations, some
276 ANGELA R. COMMITO and FELIPE ROJAS
34. H. Chanson, The Hydraulics of Roman Aqueducts:
What Do We Know? Why Should We Learn? in R. W. Babcock,
Jr., and R. Walton (eds.), World Environmental and Water Re-
sources Congress 2008 Ahupuaa, American Society of Civil En-
gineers (Reston, VA 2008) 116, at 67.
35. Hodge, Roman Aqueducts and Water Supply (supra n. 30)
347.
36. Ibid., 220.
of which amount to mere educated guesswork, is that
the Timeles aqueduct seems to have provided a volume
of water comparable to the most abundant aqueducts
serving the city of Rome.
The construction of such an enormous aqueduct for
Aphrodisias underscores a particular approach of
Roman engineers, namely, to use a standardized model
for aqueduct construction rather than to tailor each proj-
ect to local specifcations. In addition, the structure had
to be large enough to enable access by human operators,
both during construction and after for maintenance. Al-
ternatively, the engineers and patrons responsible for this
building project may have decided to build as large an
aqueduct as possible both to accommodate the possibil-
ity of increased demand in the future and to make a
strong claim of political and economic importance for
the present and future. The intentional construction of
an aqueduct with a capacity exceeding anticipated needs
may also be the result of practical considerations. Aque-
ducts were susceptible to leaks, especially where they ran
underground in long tunnels that were diffcult to access
and maintain. The anticipation of water loss in subter-
ranean segments such as the 2 km tunnel running 50 m
under the Yahsiler plateau may account in part for the
decision to transport a greater volume of water than re-
quired. In any case, certainly the Timeles was the largest
of the three Roman aqueducts to supply Aphrodisias,
which must have depended on it for full operation of the
Hadrianic Baths and, possibly, the pool stretching across
the South Agora.
Gradient
Between its source near the modern dam and the tunnel
near Kayapnar, a distance of about 11 km, the aqueduct
lost about 0.95 m in elevation per kilometer (0.095%).
37
Below Kayapnar, however, it seems to have run along
the Koca Dere and turned east to hug and then possibly
tunnel through the hillside south of the two bridges in
the Morsynus river valley. From Kayapnar to the two
bridges in the valley, the aqueduct ran at a signifcantly
steeper gradient of about 22 m per kilometer (2.2%) for
a distance of approximately 9 km. Assuming this gradi-
ent is correct, it is possible that steep chutes or drop-
shaft cascades were incorporated into the aqueduct to
help manage the fow of water. This possibility is dis-
cussed in detail below.
After crossing the bridges in the Morsynus river val-
ley, the aqueduct could have fowed at a gentler gradi-
ent of about 15 m per km (1.5%) over the distance of
approximately 7 km to Aphrodisias by following the
hills that circle the valley, before reaching the east side
of the city itself. This is the route suggested by A.
Kuznetsov using least-cost path analysis. Two suterazi,
or Ottoman water balance towers used to control
water pressure, were built along the eastern edge of the
city, suggesting that, at least in later centuries, water en-
tered the city from this direction.
Special Problem: The Koca Dere
From Kayapnar to the two bridges in the Morsynus
river valley, the Timeles aqueduct ran along the Koca
Dere at the steepest gradient of its entire route. The av-
erage gradient for this section, which covered about 9
km, is 22 m per km (2.2%). While this gradient is not
impossible (a 6 km section of the aqueduct at Carthage
ran at 28 m per km), it is certainly far above the average
slope of most Roman aqueducts, which tended to re-
main within a range of 0.55 m per km.
38
It therefore
seems likely that the aqueduct did not run at this steep
slope over the entire length of this section. Instead,
builders could have mitigated the potentially deleteri-
ous effects of rapidly fowing water by incorporating
steep chutes or drop-shaft cascades along the aqueducts
path. These structures would have allowed the conduit
to rapidly decrease in elevation by conducting the water
down vertical or near vertical declines.
Well-documented evidence of both techniques can be
found along aqueducts at Chercell, Cuicul, Cologne,
Lyon, Montjeu, and Rome.
39
In these cases, chutes and
drop shafts provided rapid changes in elevation and in
some instances were linked together in series several
kilometers in length. At Lyon, the Yzeron aqueduct in-
cluded a series of shafts at Recret that achieved an over-
all drop of 38 m along 490 m (about 78 m per km).
These shafts, each dropping in elevation by a few me-
ters, could have been part of a hydraulic stairway 2
THE AQUEDUCTS OF APHRODISIAS 277
37. Gradient estimates differ depending on whether (1) the
tunnel at aliky (E056) or the modern dam is considered to
mark the location of the source and (2) the masonry conduit in
the Koca Dere (E031) is included in the aqueduct path recon-
struction. The estimates given in this chapter work on the as-
sumption that (1) the modern dam more likely marks the source
of the aqueduct and (2) the masonry conduit in the Koca Dere
tunnel should not be included in the aqueduct path reconstruc-
tion. The possible gradients of different sections are:
aliky tunnel (E056) to Kayapnar tunnel (E009)
34 m / 14.25 km 2.39 m/km
Modern dam to Kayapnar tunnel (E009)
10.5 m / 11.04 km 0.95 m/km
Kayapnar tunnel (E009) to Koca Dere conduit (E031)
233 m / 5.69 km 40.95 m/km
Kayapnar tunnel (E009) to Gms Dere bridge (E058)
209 m / 9.49 km 22.02 m/km
Koca Dere conduit (E031) to Aphrodisias
82 m / 11.05 km 7.42 m/km
Gms Dere bridge (E058) to Aphrodisias
106 m / 7.25 km 14.62 m/km
38. Hodge, Roman Aqueducts and Water Supply (supra n. 30)
34748.
39. H. Chanson, Hydraulics of Roman Aqueducts: Steep
Chutes, Cascades, and Dropshafts, AJA 104 (2000) 4772, at
60, 67.
km long, along which some 50 shafts achieved a reduc-
tion in the elevation of the conduit by dozens of meters,
one step at a time.
40
Along the aqueduct of Chercell (Caesarea Maureta-
niae) in Algeria, a 3 km long loop along the Oued
Ilelouine valley was cut off by a bridge. To compensate
for the severe drop in elevation, a series of four steep
chutes, each followed by a drop shaft, was constructed
on the upstream side of the bridge to bring the water
level down by 12.28 m.
41
Steep declines could also be
used to lower the water level rapidly. A Hadrianic re-
structuring of the Anio Vetus in Rome shortened its
route with the addition of a new bridge, the Ponte S.
Gregorio, which compensated for the change in eleva-
tion with a steep reduction in the height of its last few
arches. Here the aqueduct lost 4.09 m in 25 m (about
164 m per km).
42
Unfortunately, the only aqueduct structure identi-
fed along the Koca Dere is the rock-cut tunnel or open
channel near the village of Kayapnar (E009) (see Fig.
36). Comparable examples of vertical cuttings made in
the rocky cliffsides of stream banks that act as bedding
for the conduit include part of the aqueduct of Cahors
where it passes through the Vers valley in France and
the aqueduct at Side on the southern coast of Turkey.
43
In these cases an open trench was excavated from the
cliff so that on the exterior side, a wall of rock remains,
part of the original cliff face. The barrel-vaulted ma-
sonry conduit of the aqueduct supplying Ephesos built
by Claudius Aristion in the second century A.D. was
laid in a rock-cut channel carved from the hills of the
Kaystros valley. According to the reconstruction, the
rock-cut channel was 1.90 m wide with a maximum
height along the rock face of 2.50 m. The internal di-
mensions of the conduit were 1.05 m wide by 1.40 m
high, making it comparable to the conduit of the Time-
les.
44
Along the Hadrianic aqueduct at Corinth, verti-
cal rock cuttings formed a secure base on top of which
the sidewalls and vault of the masonry conduit were
constructed.
45
The width of the rock-cut channel was
1.20 m, the same as the internal width of the conduit
where it was built entirely of masonry, and the interior
surfaces of the rock were coated with plaster to reduce
friction and block seepage. Thus, where possible, the
builders exploited limestone outcrops for use as a solid
and perhaps more expedient conduit foundation. Like-
wise, the aqueducts of Frjus and Amasya included a
combination of rock-cut channel and built masonry.
46
In conclusion, when it reached the Koca Dere, the
water of the Timeles aqueduct ran either in a masonry
conduit built into a rock-cut passageway along the
gorge or in the rock-cut passageway itself. The rock-cut
passageway could have been an open channel like the
structure we documented or an enclosed tunnel, or it
could have included both. It seems more likely that the
passageway was an open channel within which the en-
closed masonry conduit was built. In either case, the
conduit itself was certainly enclosed. This reconstruc-
tion corroborates our local informants description of a
now lost masonry conduit in the area, and matches the
reconstruction of the Aristion aqueduct at Ephesos.
Where a level rock-cut bedding could not be created for
the masonry conduit, builders could have instead con-
structed masonry supporting walls to serve as sub-
structures for the conduit as it ran along the gorge. Such
supporting walls have been identifed on the Hadrianic
aqueduct at Corinth, the aykavustug u aqueduct of
Tralleis, and the mid-second-century A.D. Deg ir-
mendere aqueduct at Ephesos, where supporting walls
with buttresses carried the conduit alongside steep ter-
rain.
47
Finally, although no physical evidence of such
structures has been identifed, the steep gradient of this
section of the aqueduct suggests that chutes or drop-
278 ANGELA R. COMMITO and FELIPE ROJAS
44. G. Wiplinger, Wasser fr Ephesos: Stand der Erforschung
der Wasserversorgung, in G. Wiplinger (ed.), Cura Aquarum in
Ephesos: Volumes I and II, Proceedings of the 12th International
Congress on the History of Water Management and Hydraulic
Engineering in the Mediterranean Region (Leuven 2006) 2340,
at 28.
45. Y. A. Lolos, The Hadrianic Aqueduct of Corinth, Hes-
peria 66.2 (1997) 271314, at 282.
46. Frjus: J.-M. Michel, Chapitre 4: Le canal: Caractris-
tiques et particularits, in C. Gbara, J.-M. Michel, and J.-L.
Guendon, LAqueduc Romain de Frjus. Sa Description, son His-
toire et son Environnement, Revue Archologique de Narbon-
naise, Suppl. 33 (Montpellier 2002) 12136, at 12126. Amasya:
O. Nicholson and C. Nicholson, The Aqueduct at Amasya in
Pontus, Anatolian Studies 43 (1993) 14346.
47. Corinth: Lolos, The Hadrianic Aqueduct of Corinth
(supra n. 45) 28385. Tralleis: N. O. Baykan and Y. E. Tanriver,
Water Supply System of the Ancient City Tralleis, in Wiplinger,
Cura Aquarum in Ephesos (supra n. 44) 13336, at 135. Ephesos:
Wiplinger, Wasser fr Ephesos (supra n. 44) 3233.
40. Burdy, Some Directions of Future Research (supra n.
33) 3435, fg. 5; idem, Les aqueducs romains de Lyon (supra n.
31) 12930, fg. 57.
41. Ph. Leveau and J.-L. Paillet, LAlimentation en Eau de
Caesarea de Maurtanie et LAqueduc de Cherchel (Paris 1976)
150; Hodge, Roman Aqueducts and Water Supply (supra n. 30)
146.
42. P. Aicher, Guide to the Aqueducts of Ancient Rome (Wau-
conda, Ill. 1995) 125; Hodge, Roman Aqueducts and Water Sup-
ply (supra n. 30) 146, citing T. Ashby, The Aqueducts of Ancient
Rome (Oxford 1935) 69, fg. 2, and 70 n. 1.
43. Cahors: D. Rigal, Laqueduc antique de Cahors, in R.
Bedon (ed.), Les Aqueducs de la Gaule Romaine et des Rgions
Voisines (Universit de Limoges 1997) 31130, at 31617, 328, fg.
4. Side: H. Fahlbusch, Side, in Die Wasserversorgung antiker
Stdte: Pergamon, Rechte/Verwaltung, Brunnen/Nymphen,
Brauelemente, Geschichte der Wasserversorgung 2 (Mainz am
Rhein 1987) 21821, fg. 2; K. Grewe, Antike Welt der Technik
VI: Die rmische Wasserleitung nach Side (Trkei), Antike Welt
25.2 (1994) 192203; . zis, Historical Water Schemes in
Turkey (supra n. 17) 34783.
shaft cascades were used to bring down the level of the
water rapidly.
Conduit
As it made its way from the Tavas plain to Aphrodisias,
the conduit ran through rock-cut tunnels, in trenches
excavated from the earth, over bridges, and atop sub-
structures. Different construction techniques may have
been used to meet the various conditions created by
each situation. Unfortunately, we have physical remains
of the conduit only where it ran underground in rock-
cut tunnels or trenches. However, an indication of what
the conduit looked like when it sat atop bridges and
substructures can be gleaned from comparable aque-
ducts elsewhere. Because measurements and other spec-
ifcations have been provided above and in the
catalogue, this section focuses instead on important
technical aspects of the conduit. Interesting to note,
however, is the remarkable size of the conduit, with an
average internal width of 1.50 m.
Features of the conduit are most visible at the Cig lek
Deresi, where the cross section of the conduit is visible
in the hillside (E035) (see Fig. 23). Here, builders dug a
trench in the ground, laid a rubble foundation, pro-
tected the foor with two layers of mortar, built the side-
walls and vault of the conduit, surrounded the sidewalls
and vault with rubble packing, and then reflled the
trench with dirt. The rubble foundation layer was
topped by a bedding of white mortar, laid thickest at the
center of the conduit. Coating this white mortar bed-
ding was a very thin layer, only 0.001 m thick, of pink
waterproofng mortar, itself covered by layers of pre-
cipitated calcium carbonate. Neither of these layers ex-
tended under the sidewalls of the conduit.
Both the thin layer of waterproofng mortar and the
mortar used to bond the blocks of the sidewalls and
rubble vault have a red tint, created by the use of pul-
verized terracotta or of the local iron-rich sandstone,
crushed up and added as aggregate to the lime. Silica
(SiO
2
) is a common cementing material in sandstone,
and silica along with alumina (Al
2
O
3
) are components
of pozzolans, materials with water-resistant properties.
Pozzolans were combined with lime, burned, and then
mixed with water to produce hydraulic cement, which
in turn could be combined with sand and water to make
a hydraulic lime mortar.
At no point along the aqueduct is there clear indica-
tion that crushed pottery or brick was used as a com-
ponent of its mortar. Instead, the red tint of one type of
mortar found throughout the aqueduct is likely due to
the use of the local iron-rich sandy conglomerates
prevalent throughout the region, crushed up and added
to the lime. Thus, the builders quarried locally available
sandstone for blocks to use in the conduit and bridges,
and crushed it to make hydraulic mortar. Similar use of
local resources for building materials has been noted
along the aqueducts of central and south Tunisia and
Tripolitania, which use predominately a type of mortar
that does not include pottery or tile (that is, not opus
signinum).
48
Oligocene sandstone in the Meliane valley
was used extensively in the construction of the aque-
duct supplying Carthage. It was quarried into blocks
used in both the interior and exterior of arcade con-
struction, and small fragments of this sandstone, prob-
ably detritus from block carving, were used as aggregate
for the concrete.
49
This red-tinted mortar is one of two types found
over the course of the Timeles aqueduct. Both are lime-
based with stone inclusions, but the other type is white
or white-gray. The lack of a red tint suggests the white
mortar includes neither crushed terracotta nor pulver-
ized sandstone, making it nonhydraulic. Interesting to
note, the white mortar is predominantly found in
bridges, while the pink mortar is found in both bridges
and the conduit. Obviously, the pink, hydraulic mortar
would have been preferred for the conduit to make it
waterproof, and because it was able to set even in moist
conditions, such as in a subterranean tunnel. Should
leaks occur, however, the white, nonhydraulic mortar
used in bridges and as a bedding under the conduit foor
would have allowed seeping water to drain from the
structure and evaporate, rather than freeze and crack the
masonry, thereby maintaining the stability of the struc-
tures. The results of this permeability are the amor-
phous deposits of calcium carbonate covering many of
the bridge piers along the aqueduct.
It is unclear whether the water ran in a masonry con-
duit where the aqueduct traveled deep underground in
long tunnels such as the one under the Yahsiler plateau.
The places where the conduit is exposedalong the
Kepiz ay (G001) and at Cig lek Deresi (E035) and
Yesilky (E006)are all close to the surface, indicating
that masonry conduits were built at least at the ends of
a tunnel where it was more susceptible to erosion but
providing no clue as to whether the subterranean tun-
nels themselves were equipped with masonry conduits.
Both scenarios exist along the aqueducts around Aphro-
disias. The rock-cut tunnel of the Isklar aqueduct on
the I
.
r -
i. ` . 0u
r[],
ri , I , v. 'yr[
0]-
45 i i _ j_ v. i v.u
.
[ c. 13 ]
o u [ ? ]
I j [ ? ]
r o
.
[j ? ]
.
[ ? ]
[ ?
When Tiberius Claudius Hypsikles, son of Hoplon, was
stephanephoros, the imperator Caesar, son of divine Trajan
Parthicus, grandson of divine Nerva, Trajan Hadrian Augus-
tus, pontifex maximus, holding tribunician power for the
eighth time, consul for the third time (A.D. 124) greets the
magistrates, the council and the people of Aphrodisias. As [ ..
? .. ] they, being gathered in a body, addressed ?me [ .. ? .. ]
aqueduct (accusative case) to get [ ? help .. ? .. name (accusative
case) son of Diogenes whom [ ?you had appointed as]
ambassador [ ...
Reynolds argued that the use of the phrase o
u (l. 46) suggested a body of people
addressing the emperor viva voce. She believed that this
probably occurred about A.D. 123 or 124 during
Hadrians stay in western Asia Minor; as Reynolds her-
self noted, this does not imply an actual visit to Aphro-
disias, or even Caria, for that matter. The words I
j (l. 47) make clear that this letter concerns an
aqueduct. The appearance of j instead of the
more usual ij or simply 0j is suggestive.
Reynolds proposed that the emperor was echoing the
exact wording the Aphrodisians themselves used in ap-
pealing to him. She argued that the prefx - was to
be taken strictly: for it is almost inevitable that water
would be brought down from the hills to the north/east
of Aphrodisias.
104
But if the word had any special force
at all, something beyond mere downward directionality
would be implied. There were likely already two or
102. Reynolds, New Letters (supra n. 2); SEG L 1096;
IAph2007 11.412.
103. IAph2007 11.412, lines 41end.
104. Reynolds, New Letters (supra n. 2) 18.
97. IAph2007 12.1111. On M. Ulpius Carminius Claudi-
anus, see most recently A.-V. Pont, Linscription CIG, 2782
dAphrodisias en lhonneur de M.dianus, Cahiers du Centre
Glotz (2009) 126, and also, P. J. Thoneman and F. Ertugrul, The
Carminii of Attouda, EpigAnat 38 (2005) 7586.
98. Cormack, Epigraphic Evidence for the Water-Supply
of Aphrodisias (supra n. 13) 910.
99. See Hodge, Roman Aqueducts and Water Supply (supra
n. 30) 6, with n. 5, who puts it succinctly: unlike such things as
theaters, even millionaires could not afford to fnance [an aque-
duct] single-handedly. K. M. Coleman, Exchanging Gladiators
for an Aqueduct at Aphrodisias (SEG 50.1096), Acta Classica
51 (2008) 41, lists some extraordinary exceptions, including those
attested in CIL 2.3361 and IGRP 3.804. On the funding of
Roman aqueducts, see W. Eck, Die Wasserversorgung in rmis-
chen Reich: Sozio-politische Bedingungen, Recht und Adminis-
tration, in R. Frei-Stolba and M. A. Speidel (eds.), Die
Wasserversorgung antiker Stdte. Geschichte der Wasserver-
sorgung, 2 (Mainz 1987) 7479, and also Ph. Leveau, Aqueduct
Building: Financing and Costs, in D. E. Blackman and A. T.
Hodge (eds.), Frontinus Legacy: Essays on Frontinus de aquis
urbis Romae (Ann Arbor 2001) 85101.
100. Thoneman and Ertugrul, The Carminii of Attouda
(supra n. 97), propose a foruit for him of A.D. 150180.
101. Pont, Linscription CIG, 2782 (supra n. 97) 19.
more aqueducts bringing water to the city from the hills
to the north and east; the Aphrodisians were probably
trying to emphasize that a much larger undertaking was
required to harness the waters of the Tavas plain, which
was literally across the mountain.
The letter of a year later is much better preserved.
The document reads as follows, again in Reynoldss text
and translation (modifed at lines 36-38, following
Chaniotis's interpretation):
105
27 ri i `r j.0o ,
[] l, r l,
`
, 0u r, ri
r,
30 I i. 'r 0 i
i
_ j_ i. u I 0o i j
j . ri r j
r
r-
i 0u 0u
0r 0-
u r' ro i
-
35 u, j 0 r. i r i 0
0u-
, r rl 0 i.
o
0r 0i i 0u o i 0
0o i r j u. l l '
r-
i i i u
0 u r i
j j-
40 _ r_ i_ j_ ro, _u
0u r-
. 0
In (the stephanephorate of) Claudius Hypsikles, heros. The
imperator Caesar, son of divine Trajan Parthicus, grandson of
divine Nerva, Trajan Hadrian Augustus, pontifex maximus,
holding tribunician power for the ninth time, consul for the
third time (A.D. 125) greets the magistrates, the Council and
the People of Aphrodisias. The funds which you have re-
served for the aqueduct I confrm. And since there are certain
of your citizens who say that they have been nominated for
the high priesthood when they are incapable of undertaking it,
I have referred them to you to examine whether they are able
to undertake the liturgy and are evading it, or are telling the
truth; if, however, some of them were to appear to be better
off, it is fair that they should hold the high priesthood frst. I
allow you to receive money instead of gladiatorial shows from
the high priests, and not only do I allow you this but I also
praise this proposal. The supervisors who will be chosen by
you for the water-channel will be able to get advice and help
on those matters on which they need them from my procura-
tor Pompeius Severus, to whom I have written. Farewell.
This letter sheds light not only on the dating of the
aqueduct but also on its fnancing and, more generally,
on the administrative and fscal complexities of the re-
lationship between a free city and Rome. The text of this
letter has been the subject of recent analyses with di-
vergent interpretations. According to Reynolds, the
document implies that the high priests were unwilling to
divert monies otherwise destined for gladiatorial shows
to pay for the aqueduct.
106
More recently, K. Coleman
has argued that the exact opposite was the case, and that
a subvention for the aqueduct was the solution to the
problem of the high priests unwillingness to fund glad-
iatorial shows. Coleman concluded that the imperial
priesthood in Aphrodisias preferred to be associated
with a lasting monument that was essential to the citys
daily life instead of basking in the ephemeral praise of
their fellow citizens during the day of the shows.
107
A.
Chaniotis suggested a different interpretation, consid-
ering contemporary parallels. According to Chaniotis,
Hadrians intervention resulted from disagreement in
Aphrodisias concerning two separate issues: the funding
for the aqueduct and the type of summa honoraria to
be contributed by high priests (money or shows).
108
Quite apart from local disagreements about the pre-
ferred ends of euergetism in second-century A.D.
Aphrodisias (whether games or water supply), the let-
ters provide evidence of a free citys need or desire to
secure sanction from Rome for major expenditure proj-
ectseven when only municipal resources would be
tapped to pay for such projects. The aqueduct in ques-
tion was not to be sponsored by the emperor, nor even
as a joint imperial-municipal venture, but rather exclu-
sively by the city; even so, the Aphrodisians sought
Hadrians authorization. In a brief discussion of these
inscriptions, D. Campanile has rightly called attention
to the fact that the Aphrodisians solicited imperial ap-
proval to fund their aqueduct because, among other
things, its funding involved the reallocation of monies
that were otherwise destined to pay for public spectacles
in the emperors honor.
109
Further epigraphic and literary evidence supports the
notion that by the second century A.D., cities in the
provincesfree or notsought imperial approval to
undertake urban projects at this scale. Such behavior
helps explain dedicatory inscriptions that explicitly
mention a proconsuleven when subvention for the
projects in question was furnished entirely from a citys
own resources; inclusion of the name of a proconsul
presumably refects the imperial magistrates personal
290 ANGELA R. COMMITO and FELIPE ROJAS
106. Reynolds, New Letters (supra n. 2) 19.
107. Coleman, Exchanging Gladiators for an Aqueduct
(supra n. 99). For an inscription from Late Antique Aphrodisias
mentioning lasting memory (ii ) that results
from building monuments, see C. Rouech, JRS 74 (1984) no. 8.
108. A. Chaniotis, Macht und Volk in den kaiserzeitlichen
Inschriften von Aphrodisias, in G. Urso (ed.), Popolo e potere
nel mondo antico (Pisa 2005) 4761, at 5759.
109. D. Campanile, Noterelle ai nuovi documenti da Afro-
disia, ZPE 135 (2001) 138. 105. IAph2007 11.412, lines 2741.
intervention with the emperor on behalf of the city.
110
Literary sources may also be provided; for example,
when the younger Pliny seeks Trajans advice about
public works in the cities of Bithynia, the monies in
question sometimes seem to come from municipal not
imperial or private funds.
111
The inscriptions from Aphrodisias document the pro-
cess of a free citys requesting the sanction of Rome. The
original letter from the Aphrodisians to Hadrian speci-
fed from where funds for the aqueduct would be taken
from and sought his approval. Ultimately, Hadrian con-
frmed that the Aphrodisians should use the compulsory
fee of the high priesthood to build their aqueduct. In ad-
dition to imperial approval, the emperor also promised
technical assistance that was surely required to under-
take the construction of an aqueduct more than 25 km
long, with tunnels up to 50 m deep and at least a dozen
bridges, which ranged in height from 5 m to almost 30 m.
The surveying alone would have exceeded by far the lim-
its of local specialists.
Conclusions
All this evidence indicates that the aqueduct we have
found is the same one mentioned in coins and inscrip-
tions, and that it should be dated to the second quarter of
the second century A.D. Construction of the Timeles
aqueduct probably began in the reign of Hadrian, dur-
ing or shortly after the time the letters were written.
112
Arguably, the citizenry of Aphrodisias, or rather certain
prominent Aphrodisians, decided to undertake this mas-
sive project when they heard that the emperor was com-
ing in the hope of obtaining imperial benefaction. As
mentioned above, the fact that Aphrodisias and Herakleia
both minted coins mentioning the Timeles, although they
were located in different valleys, was a source of confu-
sion for scholars attempting to elucidate local history and
topography. Robert and Robert offered a solution to this
apparent numismatic paradox by hypothesizing an aque-
duct carrying the waters of the Timeles ( Yenidere
ay) to Aphrodisias. We have found and documented
the remains of this aqueduct, thus validating the Roberts
hypothesis. But if the Timeles riverrather than the wa-
tershed of the ridge separating the Morsynus river valley
from the Tavas plaindelimited the territory between
Aphrodisias and Herakleia, there would be no paradox at
all, for this would mean that the Timeles ran along the
territory of Aphrodisias. In any case, the exact political
boundaries remain unclear, and either the Timeles river
or the mountain range separating the two valleys could
have acted as a clear physical limit between Aphrodisias
and Herakleia.
The Timeles aqueduct was the grandest aqueduct of
the city, but not the frst; the one or two bath complexes
in operation before the Hadrianic Baths would them-
selves have needed major water conveyance facilities.
But by the time of Hadrians reign, water in the
Morsynus river valley was perceived to be insuffcient
for the needs of the city and perhaps also its chora. In
the city, the Timeles aqueduct most likely fed the luxu-
rious complex known as the Hadrianic Baths, and it was
surely related to other civic monuments, such as public
fountains. It is possible that a monumental waterspout
in the shape of a ship that was found in the vicinity of
the Civic Basilica (Fig. 41) could be related to the grand
terminus of this aqueduct.
DATING, PURPOSE, AND FUNDING OF
THE AQUEDUCTS
For two and a half centuries after the foundation of
Aphrodisias in the early second century B.C., the resi-
dents of the city obtained water from wells, cisterns, and
the nearby Morsynus river.
113
The city dwellers even
benefted from seasonal springs within the city proper,
which they later collected, controlled, and cele brated in
the grand ornamental pool in the South Agora built in
the early Julio-Claudian period.
114
By the late frst cen-
tury A.D., however, water management problems
within the city had become acute, perhaps as a result of
the growing population. The existing urban infrastruc-
ture was insuffcient to handle excess seasonal waters.
The city probably suffered periodic fooding as a result
of pervasive encroachment on public areas initially des-
THE AQUEDUCTS OF APHRODISIAS 291
113. The city was founded in the early second century B.C.
when the pre-existing population around the sanctuary of
Aphrodite and that of Plarasa (modern Binge), located 15 km to
the southwest, combined at Aphrodisias. The location of the
sanctuary and the city was determined partly by the fact that the
site had a plentiful water supply. Pausanias (1.26.6) mentions a
brackish spring in connection with the sanctuary, but its exact lo-
cation is unknown; the well in the nave of the Christian church
into which the temple of Aphrodite was converted in the ffth
century A.D. was presumably made to tap a fresh water source.
114. Chaniotis, Twelve Buildings (supra n. 11) 1415, with
n. 82.
110. For a useful discussion of how an emperors generosity,
as well as that of his ministers, is variously commemorated in
provincial inscriptions, see S. Mitchell, Imperial Building in the
Eastern Roman Provinces, HSCP 91 (1987) 33365, at 34349.
111. In Plin. Ep. 10.39.13, a distinction is drawn between a
private subvention (ex privatorum pollicitationibus) and an earlier
infux of unspecifed, but presumably municipal, funds; similarly,
in Epistulae 10.37.1, the monies used by the Nicomedians to fund
an aqueduct are not associated with the emperor nor with pri-
vate individuals, and may very well have been municipal, as A.
Zuiderhoek explains in The Politics of Munifcence in the Roman
Empire (2009) 42.
112. Reynolds, New Letters (supra n. 2) 18, believes that the
Domitianic inscription IAph2007 12.314 (mentioned supra nn. 11,
20, 27) implies an aqueduct, but we would rather agree with Chan-
iotis, Twelve Buildings (supra n. 11), in thinking that it refers
merely to drains, gutter, canals, and reservoirs or holding ponds.
tined for water circulation. Eventually Adrastus, a
prominent and wealthy citizen of Aphrodisias, privately
funded city-wide measures to relieve the problem, in-
cluding the purchase of land and the building of water
holding ponds. Some of these features, rather than an
aqueduct, are most likely the ones mentioned in an
aforementioned Domitianic inscription; however, it is
also possible that Adrastus built these structures in as-
sociation with a new aqueduct constructed at the same
time.
115
The growing city certainly required a more
abundant and reliable water supply. This supply could
be met only with the construction of aqueducts tapping
more distant sources, because although fed by perennial
springs, the Morsynus rivers decreases signifcantly in
volume in the summer, and the Palamutuk ridge directs
water from the springs and surface drainage of the Baba
Dag range away from the city and toward the west.
Therefore, the Aphrodisians had to tap sources located
north and east of the city to make up for the insuffcient
local supply, and they eventually built the Timeles aque-
duct to bring water from the Tavas plain.
We have identifed and documented six separate
aqueducts, four of which supplied Aphrodisias (three
certainly in the Roman period) and two of which fed a
settlement, farmstead, or villa, or provided water for ir-
rigation. The Isklar and Derince Dere aqueducts col-
lected water north of Aphrodisias on Baba Dag , while
the Seki aqueduct tapped springs to the east of the city.
The Kavakl Dere and ren Deresi aqueducts, running
at elevations too low to have supplied the city, collected
water in the western portion of the valley. The Timeles
aqueduct brought water into the Morsynus river valley
from the Tavas plain to the southeast. The striking fact
that the most abundant water sources used today are the
same as those tapped by these aqueducts underscores an
understanding of hydrology and exploitation of natu-
ral resources on behalf of the engineers and the local res-
idents who no doubt informed them.
The different aqueducts served various purposes in
and around Aphrodisias. Several of them supplied water
to public bath complexes within the city. Presumably,
the Isklar and Seki aqueductsor at any rate, an aque-
duct or aqueducts tapping the springs on the western
slopes of Baba Dag supplied water to the one or two
bath complexes that predated the Hadrianic Baths and
are attested only in inscriptions.
116
We do not know
anything about the funding of these aqueducts.
117
Since
292 ANGELA R. COMMITO and FELIPE ROJAS
Figure 41. Waterspout found near Civic Basilica at Aphrodisias with Lionel Bier in background.
116. Supra nn. 18, 19, 26.
117. In only a few instances do we know who designed and
surveyed a Roman aqueduct. One exception is provided by a re-
markable inscription from Lambaesis, Algeria (CIL VIII 2728
ILS 5795) documenting the case of the second-century A.D. hy-
draulic engineer and legionary offcer Nonius Datus in Numidia.
Nonius was picked to survey an aqueduct in Saldae in the neigh-
boring province of Mauretania. He was called back twice, once to
correct a failed engineering job involving a tunnel; see Laporte,
Notes sur laqueduc de Saldae (supra n. 67) 74779; Grewe,
Tunnels and Canals (supra n. 59) 32933. In his correspon-
dence with Trajan while governor in Bithynia-Pontus, Pliny the
Younger describes how, after two failed attempts at building an
aqueduct, he fears he has the manpower but not the expertise to 115. Supra nn. 11, 20, 22, 27.
the frst of these complexes, the Eusebian Baths, dates to
the early Julio-Claudian period, one of the aqueducts
must have been built at least this early. Unfortunately,
we were not able to date precisely the Isklar and Seki
aqueducts, partly because the remains are scarce and not
very distinctive, and partly because local communities
have continuously reused these structures so that many
of the surviving remains are in fact a patchwork from
different periods. In addition, one or both of these aque-
ducts may have been constructed in association with
Adrastuss refurbishment of the water management sys-
tem within the city during the reign of Domitian.
The Timeles aqueduct was planned in the second
quarter of the second century A.D. to tap water sources
outside the Morsynus river valley from the southeastern
slopes of Mount Kadmos (Baba Dag ), near the town of
Herakleia (modern Vakf). It likely supplied the large
Hadrianic Baths and may also have fed other complexes
such as the so-called Theater Baths at Aphrodisias. We
know from inscriptional evidence that this aqueduct
was fnanced with funds reserved by the city and also
with monies taken from private citizens serving as
priests. We also know that other possible means of
funding, including diverting money from gladiatorial
shows, were considered but were dismissed by the em-
peror Hadrian.
118
The letters from Hadrian discussed
above in the section Historical Analysis suggest that
ambassadors from Aphrodisias asked Hadrian for f-
nancial approval and technical assistance for the con-
struction of the Timeles aqueduct.
In the later letter regarding the aqueduct, Hadrian
approves the Aphrodisians funding scheme, indicates
that the Aphrodisians should select curators for the
project, and offers the assistance of his procurator Pom-
peius Severus. The engineer and architect who surveyed
and designed the Timeles aqueduct may have been se-
lected by the procurator, by local curators with a better
knowledge of local topography and resources, or by a
contractor.
119
When, during the reign of Hadrian, the
people of Saldae in Mauretania wanted to build an aque-
duct, they requested help from the procurator of the
province, who asked the legate of Numidia to select an
engineer from the Third Legion Augusta.
120
The le-
gionary offcer and librator Nonius Datus was sent to
Saldae to survey the aqueduct, and soldiers provided
labor for the project. The Aphrodisians likely took
Hadrians offer and turned to the procurator Pompeius
Severus for recommendations. Presumably the city con-
tracted local skilled and unskilled labor for the actual
construction of the aqueduct, but the expertise required
to survey and design the technically challenging and so-
phisticated structure must have been found outside the
valley.
The Timeles aqueduct incites questions regarding the
administration and use of local resources. It seems un-
likely, for example, that subsistence reasons alone mo-
tivated the construction of this aqueduct, for both the
Morsynus river valley and the Tavas plain are quite fer-
tile. While the population of the valley in antiquity was
likely roughly comparable to what it is today,
121
pres-
ent-day farmers use only locally available water from
the springs and streams described above to manage
crops such as olives, wheat and barley, tobacco, and
grapes; no additional supply from outside the valley is
required. This situation suggests that in antiquity, too,
the water from the western slopes of Mount Kadmos
was probably suffcient for the agricultural needs of
Aphrodisias and the smaller settlements in the valley.
Although we have found clues that the Timeles aque-
duct may have supplied the residents of farmsteads, set-
tlements, villas, and agricultural sites in the valley, this
aqueduct was never equipped with the kind of elabo-
rate and substantial mechanisms for rural supply docu-
mented, for example, along aqueducts in North Africa
and at Rome discussed above in Technical Analysis,
Rural Water Supply.
122
Clearly the Timeles aqueduct
was intended to channel the vast majority of its waters
into the city itself.
Presumably the Timeles, Seki, and Isklar aqueducts
supplied more than just drinking water for the estimated
THE AQUEDUCTS OF APHRODISIAS 293
119. Fabre et al., Interdisciplinary Research on the Aque-
duct of Nmes and the Pont du Gard (supra n. 71) 7475.
120. Supra n. 117.
121. It has been estimated that in the Imperial period, the val-
ley sustained between 20,000 and 30,000 people. According to
the offcial Web page of the Karacasu municipality, http://www
.karacasu.gov.tr/Karacasu.aspx?SayfaNufus (accessed 1 Janu-
ary 2010), the valley sustains 21,980 people.
122. Wilson, Water Management and Usage in Roman North
Africa (supra n. 31) 1015; Thomas and Wilson, Water Supply
for Roman Farms (supra n. 75) 13996.
construct an aqueduct and a canal (Epistulae, 10.37, 4142, 61
62). Trajan then suggests he requisition an expert from the gov-
ernor of the Moesia Inferior, the closest province to Bithynia with
army presence. Other inscriptions attest to the existence of li-
bratores and even student engineers in the army, which required
such expertise; see, e.g., ILS 5759a, discens libratorum, from an
inscription in North Africa of the third century A.D., commem-
orating the opening of a new aqueduct. See also, P. A. Fvrier,
LArme romaine et la Construction des Aqueducs in Aqueducs
Romains, Dossiers de LArchologie no. 38 (Oct.Nov. 1979) 88
93; and Arme et Aqueducs, Journes dEtudes sur les Aque-
ducs romains, Lyon, 2628 Mai, 1977 (Paris 1983) 13340.
118. The emperor is commenting on and partly approving a
funding plan proposed by the Aphrodisians. It would seem that
even a free city such as Aphrodisias actually required the
emperors approval on major capital expenditure projects; see
Campanile, Noterelle ai nuovi documenti (supra n. 109); alter-
natively, but less likely, the correspondence might have been ini-
tiated by the Aphrodisians in an attempt to obtain imperial
benefaction. See Historical Analysis and the references there,
especially Reynolds, New Letters (supra n. 2) 14; and Coleman,
Exchanging Gladiators for an Aqueduct (supra n. 99).
10,00015,000 residents of the Imperial-period city. The
Timeles aqueduct alone may have provided 157,000 m
3
of
water each day. Perhaps the desire for a grand public bath
at a scale unprecedented for the city, and also the desire
to maintain a concentrated population in a relatively
dense urban environment, motivated members of the
Aphrodisias elite to undertake this bold project. Ulti-
mately the bath and the aqueduct itself became monu-
ments of civic cohesiveness; thus, the rather distant
Timeles river was celebrated on coins, and the occasion of
its introduction into the city was used as a chronological
reference in inscriptions.
123
Although the inscriptional
evidence suggests that the entire project and especially its
enormous cost was initially divisive, the Aphrodisians
must have remembered proudly how they managed to
survey the hilly terrain, erect numerous bridges over
steep gorges, and tunnel more than 2.5 km through the
mountain range to bring water from the Tavas plain to
their city.
The hypothesis that the main purpose of the Timeles,
Seki, and Isklar aqueducts was to supply Aphrodisias
does not preclude the limited use of their waters outside
the city. The Timeles aqueduct may have been tapped at
multiple points along its route, and the route of the
Roman Isklar aqueduct was reused multiple times after
the depopulation of the city (including iterations as the
Derince Dere pipeline and the more modern cement
channel), highlighting the fact that aqueducts were built
to supply nonurban consumers as well. More important,
these three aqueducts were not the only water lines built
in the valley. The Kavakl Dere and ren Deresi aque-
ducts ran at elevations lower than that of Aphrodisias
and therefore must have provided water to one or more
of the Roman or post-Roman farmsteads, villas, settle-
ments, and agricultural sites identifed throughout the
Morsynus river valley. The aqueducts documented by
the survey help break down the view of Roman aque-
ducts as structures built solely for the provision of large-
scale public buildings, especially those devoted to display
and recreation.
We do not yet know exactly how the water these
aqueducts carried was distributed within the city. Surely
the aqueducts that supplied water to the baths also fed
public fountains, pools, and nymphaea in Aphrodisias.
One possible terminus is a niched structure in the
northeast part of the city identifed as a nymphaeum at
the time of excavation. Though the structure contains
no pipelines or channels, its locationat an elevation
higher than the rest of the city and to the northeast,
where it seems the three Roman aqueducts entered
Aphrodisiasis optimal for water collection and distri-
bution. Judging from the extensive remains of pipes and
water facilities in Late Antique elite houses such as the
Triconch House and Atrium House, the aqueducts
eventually supplied water also to private residences. In
addition to the baths and fountains, we know from ar-
chaeological and epigraphic evidence about other in-
traurban water facilities for both storage and disposal
of water. The different pools in the city are not all re-
lated to aqueducts; some are rather reservoirs to contain
and control rainwater as well as the fow of seasonal
springs within the city.
The Hadrianic Baths associated with the Timeles
aqueduct continued to be used for at least four cen-
turies, as is attested epigraphically.
124
Presumably, even
before the city was abandoned, the aqueducts fell into
disrepair, as sinter accumulated on channel walls, and
tunnels and bridges began to collapse. Maintaining the
Timeles aqueduct was beyond the material possibilities
of the communities that inhabited the Morsynus river
valley after Late Antiquity. However, it is clear that that
some of the smaller aqueducts tapping the streams on
the western slopes of the Baba Dag range within the
Morsynus valley were in fact repaired.
During the Ottoman period, took control of the area,
these local water sources and at least some tracts of the
infrastructure were again used to supply bathhouses in
the old village of Geyre. The major change in the use and
management of water that Ottoman rule brought about
was the placement of cisterns along roads to facilitate
traffc through the region (see Fig. 40). We have docu-
mented them summarily in the catalogue below, but they
require fuller treatment by competent authorities.
The aqueducts and cisterns investigated and docu-
mented by the Aphrodisias Regional Survey emphasize
the ongoing exploitation of water as a vital natural re-
source over millennia of interaction between the regions
human population and the surrounding environment.
That the residents of Roman Aphrodisias supplied their
marble-clad city with abundant water from multiple
aqueducts is perhaps not surprising. More interesting are
the different types of waterworks in the region and the
diversity of needs they were built to accommodate.
These factors draw attention to the longevity of human-
environment interaction in the region and its changing
appearance over time. Most intriguing, however, is the
way in which the shifting forms of this interaction re-
fect much broader historical changes, as the water lines
supporting the high-density city and settled countryside
of the Hellenistic and Roman periods fell out of use, and
effort was redirected to the creation of cisterns along a
network of travel and communication that united a new
Ottoman empire.
294 ANGELA R. COMMITO and FELIPE ROJAS
124. There is epigraphic evidence confrming the maintenance
and repair of the Hadrianic Baths into the sixth century A.D. See,
e.g., eALA 74, 86, 87, and n. 106 on a sixth-century A.D. restora-
tion by Rhodopaeus of a portion of the Hadrianic Baths known
as the summer Olympian Baths. 123. See Historical Analysis.
APPENDIX
A. Calculation 1
Variables
A cross-sectional wetted area 1.50 1.30 1.95 m
2
P wetted perimeter 1.50 2(1.30) 4.10 m
R hydraulic radius A/P 1.95/4.10 0.476 m
S slope 10.5 m / 11.04 km 0.95 m/km
0.95 m/1,000 m 0.00095
(Modern dam to Kayapnar tunnel [E009])
C Chzy coeffcient 6080 smoothly plastered
concrete
(80 smoother, newer concrete)
K Kutter coeffcient 55100
(55 very rough concrete; 100 very smooth concrete)
Bazin formula for mean velocity V C (R S)
V 60 (0.476 0.00095) 1.276 m/second
(using 60 to obtain a minimum mean velocity)
V 80 (0.476 0.00095) 1.701 m/second
(using 80 to obtain a maximum mean velocity)
Q discharge AV 1.95 1.276 2.488 m
3
/second
(minimum discharge)
1.95 1.701 3.317 m
3
/second
(maximum discharge)
Manning formula for mean velocity V KR
2/3
S
1/2
V 75(0.476)
2/3
(0.00095)
1/2
1.409 m/second
Q discharge AV 1.95 1.409 2.748 m
3
/second
Discharge per day
(Bazin)
2.488 m
3
/second 86,400 second/day
214,963 m
3
/24 hours
(minimum discharge per day)
3.317 m
3
/second 86,400 second/day
286,589 m
3
/24 hours
(maximum discharge per day)
(Manning)
2.748 m
3
/second 86,400 second/day
237,427 m
3
/24 hours
B. Calculation 2
New variables
A cross-sectional wetted area 1.50 1.00 1.50 m
2
P wetted perimeter 1.50 2(1.00) 3.50 m
R hydraulic radius A/P 1.50/3.50 0.429 m
Bazin formula for mean velocity V C (R S)
V mean velocity C (R S)
60 (0.429 0.00095) 1.211 m/second
(minimum mean velocity)
80 (0.429 0.00095) 1.615 m/second
(maximum mean velocity)
Q discharge AV 1.50 1.211 1.817 m
3
/second
(minimum discharge)
1.50 1.615 2.423 m
3
/second
(maximum discharge)
Manning formula for mean velocity V KR
2/3
S
1/2
V 75(0.429)
2/3
(0.00095)
1/2
1.315 m/second
Q discharge AV 1.5 1.315 1.973 m
3
/second
Discharge per day
(Bazin)
1.817 m
3
/second 86,400 second/day
156,989 m
3
/24 hours
(minimum discharge per day)
2.423 m
3
/second 86,400 second/day
209,347 m
3
/24 hours
(maximum discharge per day)
(Manning)
1.973 m
3
/second 86,400 second/day
170,467 m
3
/24 hours
C. Calculation 3
New variables
S
1
0.00095 / 2 0.000475 (half the slope)
S
2
0.00095 2 0.0019 (twice the slope)
Bazin formula for mean velocity V C (R S)
V mean velocity C (R S
1
)
60 (0.429 0.000475) 0.856 m/second
(minimum mean velocity, half the slope)
C (R S
2
)
60 (0.429 0.0019) 1.713 m/second
(minimum mean velocity, twice the slope)
Q discharge AV 1.50 0.856 1.284 m
3
/second
(minimum discharge, half the slope)
1.50 1.713 2.570 m
3
/second
(minimum discharge, twice the slope)
Discharge per day
1.284 m
3
/second 86,400 second/day
110,938 m
3
/24 hours
(minimum discharge per day, half the slope)
2.570 m
3
/second 86,400 second/day
222,048 m
3
/24 hours
(minimum discharge per day, twice the slope)
CATALOGUE
1. Seki Aqueduct
Bridge, Yksekkemer Deresi, Seki Aqueduct (A063)
(Figs. 4, 5)
Location: The bridge crosses the Yksekkemer Deresi, a tribu-
tary of the Morsynus river. It is located 8.5 km southeast of
Aphrodisias and 1.5 km southeast of the modern village of
Seki, just northeast of the main road running through the
valley.
Elevation: 705 masl
Coordinates: Latitude: 374115.96N; longitude: 284856.10E
Preservation: The piers and arch of the bridge are well preserved
because they were incorporated into a later (probably Otto -
man) road bridge. However, the composite structure was lev-
eled and resurfaced at that time, so the masonry above the
arch and the conduit are not preserved at all. The southeast
pier is covered with calcium carbonate incrustations.
Construction: The bridge is constructed of mortared rubble
faced with petit appareil masonry. The facing blocks and
voussoirs include feldstones of quartz and sandstone, as
well as larger blocks of marble, some originally dressed for
other purposes. The facing blocks are 0.300.40 m long and
0.150.20 m high; the blocks above the projecting course of
the northwest pier are smaller than those below. The blocks
THE AQUEDUCTS OF APHRODISIAS 295
of the projecting course are up to 1.20 m long and 0.30 m
high. The voussoirs range in size from 0.30 to 0.50 m long
and 0.15 to 0.20 m wide at the extrados. The lime mortar is
pale beige with stone inclusions, most no larger than 0.05 m,
while a hard blue mortarlike material covers the joins be-
tween the large blocks of the base of the northwest pier.
Description: The Roman bridge is preserved to a height of 6 m,
a width of 1.7 m, and a length of 15 m. Large blocks form
the base of the northwest pier, founded within the soil of the
stream bank. A projecting course is visible only on the
northwest pier, 2.5 m above the ground. The arch springs
from a point two courses higher than the projecting course
and spans 5 m.
The Ottoman bridge is built against the northeast side of
the aqueduct bridge. It is constructed of mortared rubble
faced with petit appareil masonry, but the facing blocks are
smaller and more varied in size (with a maximum length of
0.60 m) than those of the aqueduct bridge, and chinking
stones are used more liberally. The bridge is 1.8 m wide and
has a projecting course that does not span the entire width
of its northwest pier because it is built in line with the pro-
jecting course of the Roman aqueduct bridge. The arch has a
slightly greater span than the arch of the aqueduct bridge,
and putlog holes are visible under the arch directly above
the projecting course and in the southwest corner of the
base of the pier.
Major Dimensions (Roman aqueduct bridge):
p. H: 6 m
W (northwest pier): 1.7 m
p. L: 15 m
Arch span: 5 m
H (ground to bottom of projecting course): 2.5 m
H (streambed to bottom of arch): 5.9 m
Photograph Numbers: D2005.04690475, D2008.13121316
Drawing Numbers: RS 28: Section looking northwest and sec-
tion looking northeast at 1:50 scale
2. Isklar Aqueduct
Tunnel, I
ncir Deresi, a deep gorge that fows into the Morsynus river
from the north. It is located almost 7 km north of Aphro-
disias and 1.5 km north of the modern town of Isklar.
Elevation: 804 masl
Coordinates: Latitude: 374600.63N; longitude: 284430.20E
Preservation: The tunnel can be followed from the north for 8
m, after which sediment infll makes it impassable. Calcare-
ous deposits from natural water percolation line all interior
surfaces of the tunnel.
Construction: The tunnel is carved into an outcrop of schist and
quartz.
Description: The tunnel lies at a height of about 2 m above the
streambed and 2.75 below the ground level of the stream
bank above. The tunnel has a roughly vaulted profle. Its
preserved height is 0.7 m, and it is 0.93 m wide at the foor
and 0.42 m wide at the ceiling. It is preserved to a length of 8
m. Niches slightly smaller than 0.10 m square line the east-
ern interior wall.
Major Dimensions:
p. H (tunnel): 0.7 m
p. W (tunnel): 0.93 m (foor); 0.42 m (ceiling)
p. L (tunnel): 8 m
Photograph Numbers: D2006.07760786
Drawing Numbers: Drawing not inked (north elevation at 1:50
scale)
Bridge, Kavakl Dere, Isklar Aqueduct (A008) (Figs. 8, 9)
Location: The remains of the bridge are located on both banks
of the Kavakl Dere, a tributary of the Morsynus river, ap-
proximately 4 km north of Aphrodisias and 1 km southwest
of the modern town of Isklar.
Elevation: 609 masl
Coordinates: Latitude: 374446.02N; longitude: 284325.08E
Preservation: One complete pier of the aqueduct bridge with
partial remains of two springing arches stands on the north
(right) bank of the stream. Other remains of the bridge in-
clude two masses of mortared rubble, one approximately 18
m from the standing pier on the opposite bank of the stream
and another about 4 m to the north of the standing pier. The
standing pier is not preserved to the height of the conduit,
and no trace of the conduit can be found in either stream
bank. The pier is pierced by a large hole as a result of col-
lapse or settling, and the pier wall under the north arch has
collapsed. The upstream (east) face of the structure is more
heavily damaged than the downstream side.
Construction: The bridge is constructed of mortared rubble
faced with petit appareil masonry. The facing blocks, most
of which are schist feldstones, are 0.100.60 m long, 0.05
0.20 m high, and 0.200.40 m deep, with extensive use of
chinking stones as small as 0.05 m long and 0.01 m high.
Courses of blocks of standard size are interchanged with
courses of smaller, more irregular stones. A few of the exter-
nal voussoirs are schist, but most are marble blocks. They
range from 0.30 to 0.40 m long, 0.100.15 m wide at the ex-
trados, and 0.200.40 m deep. Courses of four to fve blocks
span the width of the arch. The lime mortar is pale gray-tan
with densely packed stone inclusions no larger than 0.01 m.
Description: The north pier of the bridge is preserved to a
height of 7.8 m, a width of 1.65 m, and a length of 6.6 m.
The original bridge was likely more than 42 m long and 15
m high. The large blocks of the base rest on a rock outcrop.
The central arch, which may have originally spanned 9.5 m,
springs from a point 3.6 m above the base, and the three
lowest courses of voussoirs are recessed 0.15 m from the
short (south) side of the pier facing the stream. The other
voussoirs are recessed an additional 0.15 m. Above this
upper recess, the pier narrows in width by 0.10 m on either
side above a course of large facing stones. This lateral ledge
extends even to the voussoirs of the arch spanning the
stream, with the result that the lowest three courses of the
arch are wider than the rest. In contrast, the arch on the
north side of the pier has no ledges to support a framework,
but a putlog hole just above the springing of the arch fulflls
the same purpose.
Major Dimensions:
p. H: 7.8 m
W (north pier): 1.65 m
p. L: 6.6 m
Projected original H: 15 m
Projected original L: more than 42 m
Projected original H (conduit above streambed): 13.5 m
Projected original span (central arch): 9.5 m
Projected H (arch bottom from streambed): 11.5 m
Photograph Numbers: D2005.00530061, D2008.13851400
Drawing Numbers: RS 29: Section looking west and section
looking north at 1:50 scale
Bridge, Derince Dere, Isklar Aqueduct (E050) (Fig. 10)
Location: The wall is built into the east (left) bank of the Der-
ince Dere, a tributary of the Morsynus river, 0.7 km north of
Aphrodisias, and appears to be the foundation of a pier of
an aqueduct bridge.
296 ANGELA R. COMMITO and FELIPE ROJAS
Elevation: 544 masl
Coordinates: Latitude: 374305.019N; longitude:
284304.661E
Preservation: The wall is preserved to a height of 2 m, above
which it has been leveled to match the elevation of the
stream bank into which it is built.
Construction: The wall is constructed of mortared rubble faced
with petit appareil masonry, including feldstones of quartz,
schist, and sandstone varying greatly in size. The lime
mortar is tan-gray and has stone inclusions no larger than
0.01 m.
Description: The wall is preserved to a height of 2 m and a
width of 0.80 m. The bottom of the wall is anchored by a
stone larger than the others used in the wall. A construction
trench 0.70 m wide is visible on the south (downstream) side
of the wall and contains yellow-tan sandy infll that cuts
into the bedded stream deposits of the bank.
Major Dimensions:
p. H: 2 m
Th: 0.80 m
Photograph Numbers: Unknown
Drawing Numbers: None (notebook sketch only)
3. Derince Dere Aqueduct
Channel, I